Policy: Clarify whose signature should go in debian/changelog (4.4)

[debian/debian-policy.git] / policy.sgml

<!doctype debiandoc system [
<!-- include version information so we don't have to hard code it
     within the document -->
<!entity % versiondata SYSTEM "version.ent"> %versiondata;
<!-- current Debian changes file format -->
<!entity changesversion "1.8">
]>
<debiandoc>

  <book>
    <titlepag>
      <title>Debian Policy Manual</title>
      <author><qref id="authors">The Debian Policy Mailing List</qref></author>
      <version>version &version;, &date;</version>

      <abstract>
        This manual describes the policy requirements for the Debian
        distribution.  This includes the structure and
        contents of the Debian archive and several design issues of
        the operating system, as well as technical requirements that
        each package must satisfy to be included in the distribution.
      </abstract>

      <copyright>
        <copyrightsummary>
          Copyright &copy; 1996,1997,1998 Ian Jackson
          and Christian Schwarz.
        </copyrightsummary>
        <p>
          These are the copyright dates of the original Policy manual.
          Since then, this manual has been updated by many others.  No
          comprehensive collection of copyright notices for subsequent
          work exists.
        </p>

        <p>
          This manual is free software; you may redistribute it and/or
          modify it under the terms of the GNU General Public License
          as published by the Free Software Foundation; either version
          2, or (at your option) any later version.
        </p>

        <p>
          This is distributed in the hope that it will be useful, but
          <em>without any warranty</em>; without even the implied
          warranty of merchantability or fitness for a particular
          purpose.  See the GNU General Public License for more
          details.
        </p>

        <p>
          A copy of the GNU General Public License is available as
          <file>/usr/share/common-licenses/GPL</file> in the Debian
          distribution or on the World Wide Web at
          <url id="http://www.gnu.org/copyleft/gpl.html"
               name="the GNU General Public Licence">. You can also
          obtain it by writing to the Free Software Foundation, Inc.,
          51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA.
        </p>
      </copyright>
    </titlepag>

    <toc detail="sect1">

    <chapt id="scope">
      <heading>About this manual</heading>
      <sect>
        <heading>Scope</heading>
        <p>
          This manual describes the policy requirements for the Debian
          distribution. This includes the structure and
          contents of the Debian archive and several design issues of the
          operating system, as well as technical requirements that
          each package must satisfy to be included in the
          distribution.
        </p>

        <p>
          This manual also describes Debian policy as it relates to
          creating Debian packages. It is not a tutorial on how to build
          packages, nor is it exhaustive where it comes to describing
          the behavior of the packaging system. Instead, this manual
          attempts to define the interface to the package management
          system that the developers have to be conversant with.<footnote>
              Informally, the criteria used for inclusion is that the
              material meet one of the following requirements:
              <taglist compact="compact">
                <tag>Standard interfaces</tag>
                <item>
                    The material presented represents an interface to
                    the packaging system that is mandated for use, and
                    is used by, a significant number of packages, and
                    therefore should not be changed without peer
                    review. Package maintainers can then rely on this
                    interface not changing, and the package management
                    software authors need to ensure compatibility with
                    this interface definition. (Control file and
                    changelog file formats are examples.)
                </item>
                <tag>Chosen Convention</tag>
                <item>
                    If there are a number of technically viable choices
                    that can be made, but one needs to select one of
                    these options for inter-operability. The version
                    number format is one example.
                </item>
              </taglist>
              Please note that these are not mutually exclusive;
              selected conventions often become parts of standard
              interfaces.
          </footnote>
        </p>

        <p>
          The footnotes present in this manual are
          merely informative, and are not part of Debian policy itself.
        </p>

        <p>
          The appendices to this manual are not necessarily normative,
          either. Please see <ref id="pkg-scope"> for more information.
        </p>

        <p>
          In the normative part of this manual,
          the words <em>must</em>, <em>should</em> and
          <em>may</em>, and the adjectives <em>required</em>,
          <em>recommended</em> and <em>optional</em>, are used to
          distinguish the significance of the various guidelines in
          this policy document. Packages that do not conform to the
          guidelines denoted by <em>must</em> (or <em>required</em>)
          will generally not be considered acceptable for the Debian
          distribution. Non-conformance with guidelines denoted by
          <em>should</em> (or <em>recommended</em>) will generally be
          considered a bug, but will not necessarily render a package
          unsuitable for distribution. Guidelines denoted by
          <em>may</em> (or <em>optional</em>) are truly optional and
          adherence is left to the maintainer's discretion.
        </p>

        <p>
          These classifications are roughly equivalent to the bug
          severities <em>serious</em> (for <em>must</em> or
          <em>required</em> directive violations), <em>minor</em>,
          <em>normal</em> or <em>important</em>
          (for <em>should</em> or <em>recommended</em> directive
          violations) and <em>wishlist</em> (for <em>optional</em>
          items).
          <footnote>
                Compare RFC 2119.  Note, however, that these words are
                used in a different way in this document.
          </footnote>
        </p>

        <p>
          Much of the information presented in this manual will be
          useful even when building a package which is to be
          distributed in some other way or is intended for local use
          only.
        </p>

        <p>
          udebs (stripped-down binary packages used by the Debian Installer) do
          not comply with all of the requirements discussed here.  See the
          <url name="Debian Installer internals manual"
          id="http://d-i.alioth.debian.org/doc/internals/ch03.html"> for more
          information about them.
        </p>
      </sect>

      <sect>
        <heading>New versions of this document</heading>

        <p>
          This manual is distributed via the Debian package
          <package><url name="debian-policy"
              id="http://packages.debian.org/debian-policy"></package> 
          (<httpsite>packages.debian.org</httpsite> 
          <httppath>/debian-policy</httppath>).
        </p>

        <p>
          The current version of this document is also available from
          the Debian web mirrors at
          <tt><url name="/doc/debian-policy/"
                id="http://www.debian.org/doc/debian-policy/"></tt>.
          (
          <httpsite>www.debian.org</httpsite>
          <httppath>/doc/debian-policy/</httppath>)
          Also available from the same directory are several other
          formats: <file>policy.html.tar.gz</file>
          (<httppath>/doc/debian-policy/policy.html.tar.gz</httppath>),
          <file>policy.pdf.gz</file> 
          (<httppath>/doc/debian-policy/policy.pdf.gz</httppath>)
          and <file>policy.ps.gz</file>
          (<httppath>/doc/debian-policy/policy.ps.gz</httppath>).
        </p>

        <p>
          The <package>debian-policy</package> package also includes the file
          <file>upgrading-checklist.txt.gz</file> which indicates policy
          changes between versions of this document.
        </p>
      </sect>

      <sect id="authors">
        <heading>Authors and Maintainers</heading>

        <p>
          Originally called "Debian GNU/Linux Policy Manual", this
          manual was initially written in 1996 by Ian Jackson.
          It was revised on November 27th, 1996 by David A. Morris.
          Christian Schwarz added new sections on March 15th, 1997,
          and reworked/restructured it in April-July 1997.
          Christoph Lameter contributed the "Web Standard".
          Julian Gilbey largely restructured it in 2001.
        </p>

        <p>
          Since September 1998, the responsibility for the contents of
          this document lies on the <url name="debian-policy mailing list"
          id="mailto:debian-policy@lists.debian.org">. Proposals
          are discussed there and inserted into policy after a certain
          consensus is established.
          <!-- insert shameless policy-process plug here eventually -->
          The actual editing is done by a group of maintainers that have
          no editorial powers. These are the current maintainers:

          <enumlist>
            <item>Russ Allbery</item>
            <item>Bill Allombert</item>
            <item>Andrew McMillan</item>
            <item>Manoj Srivastava</item>
            <item>Colin Watson</item>
          </enumlist>
        </p>

        <p>
          While the authors of this document have tried hard to avoid
          typos and other errors, these do still occur. If you discover
          an error in this manual or if you want to give any
          comments, suggestions, or criticisms please send an email to
          the Debian Policy List,
          <email>debian-policy@lists.debian.org</email>, or submit a
          bug report against the <tt>debian-policy</tt> package.
        </p>

        <p>
          Please do not try to reach the individual authors or maintainers
          of the Policy Manual regarding changes to the Policy.
        </p>
      </sect>

      <sect id="related">
        <heading>Related documents</heading>

        <p>
          There are several other documents other than this Policy Manual
          that are necessary to fully understand some Debian policies and
          procedures.
        </p>

        <p>
          The external "sub-policy" documents are referred to in:
          <list compact="compact">
            <item><ref id="fhs"></item>
            <item><ref id="virtual_pkg"></item>
            <item><ref id="menus"></item>
            <item><ref id="perl"></item>
            <item><ref id="maintscriptprompt"></item>
            <item><ref id="emacs"></item>
          </list>
        </p>

        <p>
          In addition to those, which carry the weight of policy, there
          is the Debian Developer's Reference. This document describes
          procedures and resources for Debian developers, but it is
          <em>not</em> normative; rather, it includes things that don't
          belong in the Policy, such as best practices for developers.
        </p>

        <p>
          The Developer's Reference is available in the
          <package>developers-reference</package> package.
          It's also available from the Debian web mirrors at
          <tt><url name="/doc/developers-reference/"
                id="http://www.debian.org/doc/developers-reference/"></tt>.
        </p>

        <p>
          Finally, a <qref id="copyrightformat">specification for
          machine-readable copyright files</qref> is maintained as part of
          the <package>debian-policy</package> package using the same
          procedure as the other policy documents.  Use of this format is
          optional.
        </p>
      </sect>

      <sect id="definitions">
        <heading>Definitions</heading>

        <p>
          The following terms are used in this Policy Manual:
          <taglist>
            <tag>ASCII</tag>
            <item>
              The character encoding specified by ANSI X3.4-1986 and its
              predecessor standards, referred to in MIME as US-ASCII, and
              corresponding to an encoding in eight bits per character of
              the first 128 <url id="http://www.unicode.org/"
              name="Unicode"> characters, with the eighth bit always zero.
            </item>
            <tag>UTF-8</tag>
            <item>
              The transformation format (sometimes called encoding) of
              <url id="http://www.unicode.org/" name="Unicode"> defined by
              <url id="http://www.rfc-editor.org/rfc/rfc3629.txt"
              name="RFC 3629">.  UTF-8 has the useful property of having
              ASCII as a subset, so any text encoded in ASCII is trivially
              also valid UTF-8.
            </item>
          </taglist>
        </p>
      </sect>
    </chapt>


    <chapt id="archive">
      <heading>The Debian Archive</heading>

      <p>
        The Debian system is maintained and distributed as a
        collection of <em>packages</em>. Since there are so many of
        them (currently well over 15000), they are split into
        <em>sections</em> and given <em>priorities</em> to simplify
        the handling of them.
      </p>

      <p>
        The effort of the Debian project is to build a free operating
        system, but not every package we want to make accessible is
        <em>free</em> in our sense (see the Debian Free Software
        Guidelines, below), or may be imported/exported without
        restrictions. Thus, the archive is split into areas<footnote>
          The Debian archive software uses the term "component" internally
          and in the Release file format to refer to the division of an
          archive.  The Debian Social Contract simply refers to "areas."
          This document uses terminology similar to the Social Contract.
        </footnote> based on their licenses and other restrictions.
      </p>

      <p>
        The aims of this are:

        <list compact="compact">
          <item>to allow us to make as much software available as we can</item>
          <item>to allow us to encourage everyone to write free software,
                and</item>
          <item>to allow us to make it easy for people to produce
                CD-ROMs of our system without violating any licenses,
                import/export restrictions, or any other laws.</item>
        </list>
      </p>

      <p>
        The <em>main</em> archive area forms the <em>Debian distribution</em>.
      </p>

      <p>
        Packages in the other archive areas (<tt>contrib</tt>,
        <tt>non-free</tt>) are not considered to be part of the Debian
        distribution, although we support their use and provide
        infrastructure for them (such as our bug-tracking system and
        mailing lists). This Debian Policy Manual applies to these
        packages as well.
      </p>

      <sect id="dfsg">
        <heading>The Debian Free Software Guidelines</heading>
        <p>
          The Debian Free Software Guidelines (DFSG) form our
          definition of "free software".  These are:
            <taglist>
              <tag>1. Free Redistribution
              </tag>
              <item>
                  The license of a Debian component may not restrict any
                  party from selling or giving away the software as a
                  component of an aggregate software distribution
                  containing programs from several different
                  sources. The license may not require a royalty or
                  other fee for such sale.
              </item>
              <tag>2. Source Code
              </tag>
              <item>
                  The program must include source code, and must allow
                  distribution in source code as well as compiled form.
              </item>
              <tag>3. Derived Works
              </tag>
              <item>
                  The license must allow modifications and derived
                  works, and must allow them to be distributed under the
                  same terms as the license of the original software.
              </item>
              <tag>4. Integrity of The Author's Source Code
              </tag>
              <item>
                  The license may restrict source-code from being
                  distributed in modified form <em>only</em> if the
                  license allows the distribution of "patch files"
                  with the source code for the purpose of modifying the
                  program at build time. The license must explicitly
                  permit distribution of software built from modified
                  source code. The license may require derived works to
                  carry a different name or version number from the
                  original software.  (This is a compromise. The Debian
                  Project encourages all authors to not restrict any
                  files, source or binary, from being modified.)
              </item>
              <tag>5. No Discrimination Against Persons or Groups
              </tag>
              <item>
                  The license must not discriminate against any person
                  or group of persons.
              </item>
              <tag>6. No Discrimination Against Fields of Endeavor
              </tag>
              <item>
                  The license must not restrict anyone from making use
                  of the program in a specific field of endeavor. For
                  example, it may not restrict the program from being
                  used in a business, or from being used for genetic
                  research.
              </item>
              <tag>7. Distribution of License
              </tag>
              <item>
                  The rights attached to the program must apply to all
                  to whom the program is redistributed without the need
                  for execution of an additional license by those
                  parties.
              </item>
              <tag>8. License Must Not Be Specific to Debian
              </tag>
              <item>
                  The rights attached to the program must not depend on
                  the program's being part of a Debian system. If the
                  program is extracted from Debian and used or
                  distributed without Debian but otherwise within the
                  terms of the program's license, all parties to whom
                  the program is redistributed must have the same
                  rights as those that are granted in conjunction with
                  the Debian system.
              </item>
              <tag>9. License Must Not Contaminate Other Software
              </tag>
              <item>
                  The license must not place restrictions on other
                  software that is distributed along with the licensed
                  software. For example, the license must not insist
                  that all other programs distributed on the same medium
                  must be free software.
              </item>
              <tag>10. Example Licenses
              </tag>
              <item>
                  The "GPL," "BSD," and "Artistic" licenses are examples of
                  licenses that we consider <em>free</em>.
              </item>
            </taglist>
        </p>
      </sect>

      <sect id="sections">
        <heading>Archive areas</heading>

        <sect1 id="main">
          <heading>The main archive area</heading>

          <p>
            The <em>main</em> archive area comprises the Debian
            distribution.  Only the packages in this area are considered
            part of the distribution.  None of the packages in
            the <em>main</em> archive area require software outside of
            that area to function.  Anyone may use, share, modify and
            redistribute the packages in this archive area
            freely<footnote>
              See <url id="http://www.debian.org/intro/free"
                       name="What Does Free Mean?"> for
              more about what we mean by free software.
            </footnote>.
          </p>

          <p>
            Every package in <em>main</em> must comply with the DFSG
            (Debian Free Software Guidelines).
          </p>

          <p>
            In addition, the packages in <em>main</em>
            <list compact="compact">
              <item>
                  must not require or recommend a package outside
                  of <em>main</em> for compilation or execution (thus, the
                  package must not declare a "Pre-Depends", "Depends",
                  "Recommends", "Build-Depends", or "Build-Depends-Indep"
                  relationship on a non-<em>main</em> package),
              </item>
              <item>
                  must not be so buggy that we refuse to support them,
                  and
              </item>
              <item>
                  must meet all policy requirements presented in this
                  manual.
              </item>
            </list>
          </p>

        </sect1>

        <sect1 id="contrib">
          <heading>The contrib archive area</heading>

          <p>
            The <em>contrib</em> archive area contains supplemental
            packages intended to work with the Debian distribution, but
            which require software outside of the distribution to either
            build or function.
          </p>

          <p>
            Every package in <em>contrib</em> must comply with the DFSG.
          </p>

          <p>
            In addition, the packages in <em>contrib</em>
            <list compact="compact">
              <item>
                  must not be so buggy that we refuse to support them,
                  and
              </item>
              <item>
                  must meet all policy requirements presented in this
                  manual.
              </item>
            </list>
          </p>

          <p>
            Examples of packages which would be included in
            <em>contrib</em> are:
            <list compact="compact">
              <item>
                  free packages which require <em>contrib</em>,
                  <em>non-free</em> packages or packages which are not
                  in our archive at all for compilation or execution,
                  and
              </item>
              <item>
                  wrapper packages or other sorts of free accessories for
                  non-free programs.
              </item>
            </list>
          </p>
        </sect1>

        <sect1 id="non-free">
          <heading>The non-free archive area</heading>

          <p>
            The <em>non-free</em> archive area contains supplemental
            packages intended to work with the Debian distribution that do
            not comply with the DFSG or have other problems that make
            their distribution problematic.  They may not comply with all
            of the policy requirements in this manual due to restrictions
            on modifications or other limitations.
          </p>

          <p>
            Packages must be placed in <em>non-free</em> if they are
            not compliant with the DFSG or are encumbered by patents
            or other legal issues that make their distribution
            problematic.
          </p>

          <p>
            In addition, the packages in <em>non-free</em>
            <list compact="compact">
              <item>
                  must not be so buggy that we refuse to support them,
                  and
              </item>
              <item>
                  must meet all policy requirements presented in this
                  manual that it is possible for them to meet.
                  <footnote>
                      It is possible that there are policy
                      requirements which the package is unable to
                      meet, for example, if the source is
                      unavailable.  These situations will need to be
                      handled on a case-by-case basis.
                  </footnote>
              </item>
            </list>
          </p>
        </sect1>

      </sect>

      <sect id="pkgcopyright">
        <heading>Copyright considerations</heading>

        <p>
          Every package must be accompanied by a verbatim copy of its
          copyright information and distribution license in the file
          <file>/usr/share/doc/<var>package</var>/copyright</file>
          (see <ref id="copyrightfile"> for further details).
        </p>

        <p>
          We reserve the right to restrict files from being included
          anywhere in our archives if
          <list compact="compact">
            <item>
                  their use or distribution would break a law,
            </item>
            <item>
                  there is an ethical conflict in their distribution or
                  use,
            </item>
            <item>
                  we would have to sign a license for them, or
            </item>
            <item>
                  their distribution would conflict with other project
                  policies.
            </item>
          </list>
        </p>

        <p>
          Programs whose authors encourage the user to make
          donations are fine for the main distribution, provided
          that the authors do not claim that not donating is
          immoral, unethical, illegal or something similar; in such
          a case they must go in <em>non-free</em>.
        </p>

        <p>
          Packages whose copyright permission notices (or patent
          problems) do not even allow redistribution of binaries
          only, and where no special permission has been obtained,
          must not be placed on the Debian FTP site and its mirrors
          at all.
        </p>

        <p>
          Note that under international copyright law (this applies
          in the United States, too), <em>no</em> distribution or
          modification of a work is allowed without an explicit
          notice saying so.  Therefore a program without a copyright
          notice <em>is</em> copyrighted and you may not do anything
          to it without risking being sued! Likewise if a program
          has a copyright notice but no statement saying what is
          permitted then nothing is permitted.
        </p>

        <p>
          Many authors are unaware of the problems that restrictive
          copyrights (or lack of copyright notices) can cause for
          the users of their supposedly-free software.  It is often
          worthwhile contacting such authors diplomatically to ask
          them to modify their license terms. However, this can be a
          politically difficult thing to do and you should ask for
          advice on the <tt>debian-legal</tt> mailing list first, as
          explained below.
        </p>

        <p>
          When in doubt about a copyright, send mail to
          <email>debian-legal@lists.debian.org</email>.  Be prepared
          to provide us with the copyright statement.  Software
          covered by the GPL, public domain software and BSD-like
          copyrights are safe; be wary of the phrases "commercial
          use prohibited" and "distribution restricted".
        </p>
      </sect>

      <sect id="subsections">
        <heading>Sections</heading>

        <p>
          The packages in the archive areas <em>main</em>,
          <em>contrib</em> and <em>non-free</em> are grouped further into
          <em>sections</em> to simplify handling.
        </p>

        <p>
          The archive area and section for each package should be
          specified in the package's <tt>Section</tt> control record (see
          <ref id="f-Section">).  However, the maintainer of the Debian
          archive may override this selection to ensure the consistency of
          the Debian distribution.  The <tt>Section</tt> field should be
          of the form:
          <list compact="compact">
            <item>
                  <em>section</em> if the package is in the
                  <em>main</em> archive area,
            </item>
            <item>
                  <em>area/section</em> if the package is in
                  the <em>contrib</em> or <em>non-free</em>
                  archive areas.
            </item>
          </list>
        </p>

        <p>
          The Debian archive maintainers provide the authoritative
          list of sections.  At present, they are:
admin,
cli-mono,
comm,
database,
debug,
devel,
doc,
editors,
education,
electronics,
embedded,
fonts,
games,
gnome,
gnu-r,
gnustep,
graphics,
hamradio,
haskell,
httpd,
interpreters,
introspection,
java,
kde,
kernel,
libdevel,
libs,
lisp,
localization,
mail,
math,
metapackages,
misc,
net,
news,
ocaml,
oldlibs,
otherosfs,
perl,
php,
python,
ruby,
science,
shells,
sound,
tasks,
tex,
text,
utils,
vcs,
video,
web,
x11,
xfce,
zope.
          The additional section <em>debian-installer</em>
          contains special packages used by the installer and is not used
          for normal Debian packages.
        </p>

        <p>
          For more information about the sections and their definitions,
          see the <url id="http://packages.debian.org/unstable/"
                       name="list of sections in unstable">.
        </p>
      </sect>

      <sect id="priorities">
        <heading>Priorities</heading>

        <p>
          Each package should have a <em>priority</em> value, which is
          included in the package's <em>control record</em>
          (see <ref id="f-Priority">).
          This information is used by the Debian package management tools to
          separate high-priority packages from less-important packages.
        </p>

        <p>
          The following <em>priority levels</em> are recognized by the
          Debian package management tools.
          <taglist>
            <tag><tt>required</tt></tag>
            <item>
                Packages which are necessary for the proper
                functioning of the system (usually, this means that
                dpkg functionality depends on these packages).
                Removing a <tt>required</tt> package may cause your
                system to become totally broken and you may not even
                be able to use <prgn>dpkg</prgn> to put things back,
                so only do so if you know what you are doing.  Systems
                with only the <tt>required</tt> packages are probably
                unusable, but they do have enough functionality to
                allow the sysadmin to boot and install more software.
            </item>
            <tag><tt>important</tt></tag>
            <item>
                Important programs, including those which one would
                expect to find on any Unix-like system.  If the
                expectation is that an experienced Unix person who
                found it missing would say "What on earth is going on,
                where is <prgn>foo</prgn>?", it must be an
                <tt>important</tt> package.<footnote>
                    This is an important criterion because we are
                    trying to produce, amongst other things, a free
                    Unix.
                </footnote>
                Other packages without which the system will not run
                well or be usable must also have priority
                <tt>important</tt>.  This does
                <em>not</em> include Emacs, the X Window System, TeX
                or any other large applications.  The
                <tt>important</tt> packages are just a bare minimum of
                commonly-expected and necessary tools.
            </item>
            <tag><tt>standard</tt></tag>
            <item>
                These packages provide a reasonably small but not too
                limited character-mode system.  This is what will be
                installed by default if the user doesn't select anything
                else.  It doesn't include many large applications.
            </item>
            <tag><tt>optional</tt></tag>
            <item>
                (In a sense everything that isn't required is
                optional, but that's not what is meant here.) This is
                all the software that you might reasonably want to
                install if you didn't know what it was and don't have
                specialized requirements. This is a much larger system
                and includes the X Window System, a full TeX
                distribution, and many applications. Note that
                optional packages should not conflict with each other.
            </item>
            <tag><tt>extra</tt></tag>
            <item>
                This contains all packages that conflict with others
                with required, important, standard or optional
                priorities, or are only likely to be useful if you
                already know what they are or have specialized
                requirements (such as packages containing only detached
                debugging symbols).
            </item>
          </taglist>
        </p>

        <p>
          Packages must not depend on packages with lower priority
          values (excluding build-time dependencies).  In order to
          ensure this, the priorities of one or more packages may need
          to be adjusted.
        </p>
      </sect>

    </chapt>


    <chapt id="binary">
      <heading>Binary packages</heading>

      <p>
        The Debian distribution is based on the Debian
        package management system, called <prgn>dpkg</prgn>. Thus,
        all packages in the Debian distribution must be provided
        in the <tt>.deb</tt> file format.
      </p>

      <p>
        A <tt>.deb</tt> package contains two sets of files: a set of files
        to install on the system when the package is installed, and a set
        of files that provide additional metadata about the package or
        which are executed when the package is installed or removed.  This
        second set of files is called <em>control information files</em>.
        Among those files are the package maintainer scripts
        and <file>control</file>, the <qref id="binarycontrolfiles">binary
        package control file</qref> that contains the control fields for
        the package.  Other control information files include
        the <qref id="sharedlibs-symbols"><file>symbols</file> file</qref>
        or <qref id="sharedlibs-shlibdeps"><file>shlibs</file> file</qref>
        used to store shared library dependency information and
        the <file>conffiles</file> file that lists the package's
        configuration files (described in <ref id="config-files">).
      </p>

      <p>
        There is unfortunately a collision of terminology here between
        control information files and files in the Debian control file
        format.  Throughout this document, a <em>control file</em> refers
        to a file in the Debian control file format.  These files are
        documented in <ref id="controlfields">.  Only files referred to
        specifically as <em>control information files</em> are the files
        included in the control information file member of
        the <file>.deb</file> file format used by binary packages.  Most
        control information files are not in the Debian control file
        format.
      </p>

      <sect>
        <heading>The package name</heading>

        <p>
          Every package must have a name that's unique within the Debian
          archive.
        </p>

        <p>
          The package name is included in the control field
          <tt>Package</tt>, the format of which is described
          in <ref id="f-Package">.
          The package name is also included as a part of the file name
          of the <tt>.deb</tt> file.
        </p>
      </sect>

      <sect id="versions">
        <heading>The version of a package</heading>

        <p>
          Every package has a version number recorded in its
          <tt>Version</tt> control file field, described in
          <ref id="f-Version">.
        </p>

        <p>
          The package management system imposes an ordering on version
          numbers, so that it can tell whether packages are being up- or
          downgraded and so that package system front end applications
          can tell whether a package it finds available is newer than
          the one installed on the system.  The version number format
          has the most significant parts (as far as comparison is
          concerned) at the beginning.
        </p>

        <p>
          If an upstream package has problematic version numbers they
          should be converted to a sane form for use in the
          <tt>Version</tt> field.
        </p>

        <sect1>
          <heading>Version numbers based on dates</heading>

          <p>
            In general, Debian packages should use the same version
            numbers as the upstream sources.  However, upstream version
            numbers based on some date formats (sometimes used for
            development or "snapshot" releases) will not be ordered
            correctly by the package management software.  For
            example, <prgn>dpkg</prgn> will consider "96May01" to be
            greater than "96Dec24".
          </p>

          <p>
            To prevent having to use epochs for every new upstream
            version, the date-based portion of any upstream version number
            should be given in a way that sorts correctly: four-digit year
            first, followed by a two-digit numeric month, followed by a
            two-digit numeric date, possibly with punctuation between the
            components.
          </p>

          <p>
            Native Debian packages (i.e., packages which have been written
            especially for Debian) whose version numbers include dates
            should also follow these rules.  If punctuation is desired
            between the date components, remember that hyphen (<tt>-</tt>)
            cannot be used in native package versions.  Period
            (<tt>.</tt>) is normally a good choice.
          </p>
        </sect1>

      </sect>

      <sect id="maintainer">
        <heading>The maintainer of a package</heading>

        <p>
          Every package must have a maintainer, except for orphaned
          packages as described below.  The maintainer may be one person
          or a group of people reachable from a common email address, such
          as a mailing list.  The maintainer is responsible for
          maintaining the Debian packaging files, evaluating and
          responding appropriately to reported bugs, uploading new
          versions of the package (either directly or through a sponsor),
          ensuring that the package is placed in the appropriate archive
          area and included in Debian releases as appropriate for the
          stability and utility of the package, and requesting removal of
          the package from the Debian distribution if it is no longer
          useful or maintainable.
        </p>

        <p>
          The maintainer must be specified in the <tt>Maintainer</tt>
          control field with their correct name and a working email
          address.  The email address given in the <tt>Maintainer</tt>
          control field must accept mail from those role accounts in
          Debian used to send automated mails regarding the package.  This
          includes non-spam mail from the bug-tracking system, all mail
          from the Debian archive maintenance software, and other role
          accounts or automated processes that are commonly agreed on by
          the project.<footnote>
            A sample implementation of such a whitelist written for the
            Mailman mailing list management software is used for mailing
            lists hosted by alioth.debian.org.
          </footnote>
          If one person or team maintains several packages, they should
          use the same form of their name and email address in
          the <tt>Maintainer</tt> fields of those packages.
        </p>

        <p>
          The format of the <tt>Maintainer</tt> control field is
          described in <ref id="f-Maintainer">.
        </p>

        <p>
          If the maintainer of the package is a team of people with a
          shared email address, the <tt>Uploaders</tt> control field must
          be present and must contain at least one human with their
          personal email address.  See <ref id="f-Uploaders"> for the
          syntax of that field.
        </p>

        <p>
          An orphaned package is one with no current maintainer.  Orphaned
          packages should have their <tt>Maintainer</tt> control field set
          to <tt>Debian QA Group &lt;packages@qa.debian.org&gt;</tt>.
          These packages are considered maintained by the Debian project
          as a whole until someone else volunteers to take over
          maintenance.<footnote>
            The detailed procedure for gracefully orphaning a package can
            be found in the Debian Developer's Reference
            (see <ref id="related">).
          </footnote>
        </p>
      </sect>

      <sect id="descriptions">
        <heading>The description of a package</heading>

        <p>
          Every Debian package must have a <tt>Description</tt> control
          field which contains a synopsis and extended description of the
          package.  Technical information about the format of the
          <tt>Description</tt> field is in <ref id="f-Description">.
        </p>

        <p>
          The description should describe the package (the program) to a
          user (system administrator) who has never met it before so that
          they have enough information to decide whether they want to
          install it. This description should not just be copied verbatim
          from the program's documentation.
        </p>

        <p>
          Put important information first, both in the synopsis and
          extended description.  Sometimes only the first part of the
          synopsis or of the description will be displayed.  You can
          assume that there will usually be a way to see the whole
          extended description.
        </p>

        <p>
          The description should also give information about the
          significant dependencies and conflicts between this package
          and others, so that the user knows why these dependencies and
          conflicts have been declared.
        </p>

        <p>
          Instructions for configuring or using the package should
          not be included (that is what installation scripts,
          manual pages, info files, etc., are for).  Copyright
          statements and other administrivia should not be included
          either (that is what the copyright file is for).
        </p>

        <sect1 id="synopsis"><heading>The single line synopsis</heading>

          <p>
            The single line synopsis should be kept brief - certainly
            under 80 characters.
          </p>

          <p>
            Do not include the package name in the synopsis line.  The
            display software knows how to display this already, and you
            do not need to state it.  Remember that in many situations
            the user may only see the synopsis line - make it as
            informative as you can.
          </p>

        </sect1>

        <sect1 id="extendeddesc"><heading>The extended description</heading>

          <p>
            Do not try to continue the single line synopsis into the
            extended description.  This will not work correctly when
            the full description is displayed, and makes no sense
            where only the summary (the single line synopsis) is
            available.
          </p>

          <p>
            The extended description should describe what the package
            does and how it relates to the rest of the system (in terms
            of, for example, which subsystem it is which part of).
          </p>

          <p>
            The description field needs to make sense to anyone, even
            people who have no idea about any of the things the
            package deals with.<footnote>
                The blurb that comes with a program in its
                announcements and/or <prgn>README</prgn> files is
                rarely suitable for use in a description.  It is
                usually aimed at people who are already in the
                community where the package is used.
            </footnote>
          </p>

        </sect1>

      </sect>

      <sect id="dependencies">
        <heading>Dependencies</heading>

        <p>
          Every package must specify the dependency information
          about other packages that are required for the first to
          work correctly.
        </p>

        <p>
          For example, a dependency entry must be provided for any
          shared libraries required by a dynamically-linked executable
          binary in a package.
        </p>

        <p>
          Packages are not required to declare any dependencies they
          have on other packages which are marked <tt>Essential</tt>
          (see below), and should not do so unless they depend on a
          particular version of that package.<footnote>
            <p>
              Essential is needed in part to avoid unresolvable dependency
              loops on upgrade.  If packages add unnecessary dependencies
              on packages in this set, the chances that there
              <strong>will</strong> be an unresolvable dependency loop
              caused by forcing these Essential packages to be configured
              first before they need to be is greatly increased.  It also
              increases the chances that frontends will be unable to
              <strong>calculate</strong> an upgrade path, even if one
              exists.
            </p>
            <p>
              Also, functionality is rarely ever removed from the
              Essential set, but <em>packages</em> have been removed from
              the Essential set when the functionality moved to a
              different package. So depending on these packages <em>just
              in case</em> they stop being essential does way more harm
              than good.
            </p>
          </footnote>
        </p>

        <p>
          Sometimes, unpacking one package requires that another package
          be first unpacked <em>and</em> configured.  In this case, the
          depending package must specify this dependency in
          the <tt>Pre-Depends</tt> control field.
        </p>

        <p>
          You should not specify a <tt>Pre-Depends</tt> entry for a
          package before this has been discussed on the
          <tt>debian-devel</tt> mailing list and a consensus about
          doing that has been reached.
        </p>

        <p>
          The format of the package interrelationship control fields is
          described in <ref id="relationships">.
        </p>
      </sect>

      <sect id="virtual_pkg">
        <heading>Virtual packages</heading>

        <p>
          Sometimes, there are several packages which offer
          more-or-less the same functionality. In this case, it's
          useful to define a <em>virtual package</em> whose name
          describes that common functionality.  (The virtual
          packages only exist logically, not physically; that's why
          they are called <em>virtual</em>.) The packages with this
          particular function will then <em>provide</em> the virtual
          package. Thus, any other package requiring that function
          can simply depend on the virtual package without having to
          specify all possible packages individually.
        </p>

        <p>
          All packages should use virtual package names where
          appropriate, and arrange to create new ones if necessary.
          They should not use virtual package names (except privately,
          amongst a cooperating group of packages) unless they have
          been agreed upon and appear in the list of virtual package
          names. (See also <ref id="virtual">)
        </p>

        <p>
          The latest version of the authoritative list of virtual
          package names can be found in the <tt>debian-policy</tt> package.
          It is also available from the Debian web mirrors at
          <tt><url name="/doc/packaging-manuals/virtual-package-names-list.txt"
                id="http://www.debian.org/doc/packaging-manuals/virtual-package-names-list.txt"></tt>.
        </p>

        <p>
          The procedure for updating the list is described in the preface
          to the list.
        </p>

      </sect>

      <sect>
        <heading>Base system</heading>

        <p>
          The <tt>base system</tt> is a minimum subset of the Debian
          system that is installed before everything else
          on a new system. Only very few packages are allowed to form
          part of the base system, in order to keep the required disk
          usage very small.
        </p>

        <p>
          The base system consists of all those packages with priority
          <tt>required</tt> or <tt>important</tt>. Many of them will
          be tagged <tt>essential</tt> (see below).
        </p>
      </sect>

      <sect>
        <heading>Essential packages</heading>

        <p>
          Essential is defined as the minimal set of functionality that
          must be available and usable on the system at all times, even
          when packages are in the "Unpacked" state.
          Packages are tagged <tt>essential</tt> for a system using the
          <tt>Essential</tt> control field.  The format of the
          <tt>Essential</tt> control field is described in <ref
          id="f-Essential">.
        </p>

        <p>
          Since these packages cannot be easily removed (one has to
          specify an extra <em>force option</em> to
          <prgn>dpkg</prgn> to do so), this flag must not be used
          unless absolutely necessary.  A shared library package
          must not be tagged <tt>essential</tt>; dependencies will
          prevent its premature removal, and we need to be able to
          remove it when it has been superseded.
        </p>

        <p>
          Since dpkg will not prevent upgrading of other packages
          while an <tt>essential</tt> package is in an unconfigured
            state, all <tt>essential</tt> packages must supply all of
            their core functionality even when unconfigured. If the
            package cannot satisfy this requirement it must not be
            tagged as essential, and any packages depending on this
            package must instead have explicit dependency fields as
            appropriate.
        </p>

        <p>
          Maintainers should take great care in adding any programs,
          interfaces, or functionality to <tt>essential</tt> packages.
          Packages may assume that functionality provided by
          <tt>essential</tt> packages is always available without
          declaring explicit dependencies, which means that removing
          functionality from the Essential set is very difficult and is
          almost never done.  Any capability added to an
          <tt>essential</tt> package therefore creates an obligation to
          support that capability as part of the Essential set in
          perpetuity.
        </p>

        <p>
          You must not tag any packages <tt>essential</tt> before
          this has been discussed on the <tt>debian-devel</tt>
          mailing list and a consensus about doing that has been
          reached.
        </p>
      </sect>

      <sect id="maintscripts">
        <heading>Maintainer Scripts</heading>

        <p>
          The package installation scripts should avoid producing
          output which is unnecessary for the user to see and
          should rely on <prgn>dpkg</prgn> to stave off boredom on
          the part of a user installing many packages. This means,
          amongst other things, not passing the <tt>--verbose</tt>
          option to <prgn>update-alternatives</prgn>.
        </p>

        <p>
          Errors which occur during the execution of an installation
          script must be checked and the installation must not
          continue after an error.
        </p>

        <p>
          Note that in general <ref id="scripts"> applies to package
          maintainer scripts, too.
        </p>

        <p>
          You should not use <prgn>dpkg-divert</prgn> on a file belonging
          to another package without consulting the maintainer of that
          package first.  When adding or removing diversions, package
          maintainer scripts must provide the <tt>--package</tt> flag
          to <prgn>dpkg-divert</prgn> and must not use <tt>--local</tt>.
        </p>

        <p>
          All packages which supply an instance of a common command
          name (or, in general, filename) should generally use
          <prgn>update-alternatives</prgn>, so that they may be
          installed together.  If <prgn>update-alternatives</prgn>
          is not used, then each package must use
          <tt>Conflicts</tt> to ensure that other packages are
          removed.  (In this case, it may be appropriate to
          specify a conflict against earlier versions of something
          that previously did not use
          <prgn>update-alternatives</prgn>; this is an exception to
          the usual rule that versioned conflicts should be
          avoided.)
        </p>

        <sect1 id="maintscriptprompt">
          <heading>Prompting in maintainer scripts</heading>
          <p>
            Package maintainer scripts may prompt the user if
            necessary. Prompting must be done by communicating
            through a program, such as <prgn>debconf</prgn>, which
            conforms to the Debian Configuration Management
            Specification, version 2 or higher.
          </p>

          <p>
            Packages which are essential, or which are dependencies of
            essential packages, may fall back on another prompting method
            if no such interface is available when they are executed.
          </p>

          <p>
            The Debian Configuration Management Specification is included
            in the <file>debconf_specification</file> files in the
            <package>debian-policy</package> package.
            It is also available from the Debian web mirrors at
            <tt><url name="/doc/packaging-manuals/debconf_specification.html"
                id="http://www.debian.org/doc/packaging-manuals/debconf_specification.html"></tt>.
          </p>

          <p>
            Packages which use the Debian Configuration Management
            Specification may contain the additional control information
            files <file>config</file>
            and <file>templates</file>.  <file>config</file> is an
            additional maintainer script used for package configuration,
            and <file>templates</file> contains templates used for user
            prompting.  The <prgn>config</prgn> script might be run before
            the <prgn>preinst</prgn> script and before the package is
            unpacked or any of its dependencies or pre-dependencies are
            satisfied.  Therefore it must work using only the tools
            present in <em>essential</em> packages.<footnote>
                  <package>Debconf</package> or another tool that
                  implements the Debian Configuration Management
                  Specification will also be installed, and any
                  versioned dependencies on it will be satisfied
                  before preconfiguration begins.
            </footnote>
          </p>

          <p>
            Packages which use the Debian Configuration Management
            Specification must allow for translation of their user-visible
            messages by using a gettext-based system such as the one
            provided by the <package>po-debconf</package> package.
          </p>

          <p>
            Packages should try to minimize the amount of prompting
            they need to do, and they should ensure that the user
            will only ever be asked each question once.  This means
            that packages should try to use appropriate shared
            configuration files (such as <file>/etc/papersize</file> and
            <file>/etc/news/server</file>), and shared
            <package>debconf</package> variables rather than each
            prompting for their own list of required pieces of
            information.
          </p>

          <p>
            It also means that an upgrade should not ask the same
            questions again, unless the user has used
            <tt>dpkg --purge</tt> to remove the package's configuration.
            The answers to configuration questions should be stored in an
            appropriate place in <file>/etc</file> so that the user can
            modify them, and how this has been done should be
            documented.
          </p>

          <p>
            If a package has a vitally important piece of
            information to pass to the user (such as "don't run me
            as I am, you must edit the following configuration files
            first or you risk your system emitting badly-formatted
            messages"), it should display this in the
            <prgn>config</prgn> or <prgn>postinst</prgn> script and
            prompt the user to hit return to acknowledge the
            message.  Copyright messages do not count as vitally
            important (they belong in
            <file>/usr/share/doc/<var>package</var>/copyright</file>);
            neither do instructions on how to use a program (these
            should be in on-line documentation, where all the users
            can see them).
          </p>

          <p>
            Any necessary prompting should almost always be confined
            to the <prgn>config</prgn> or <prgn>postinst</prgn>
            script. If it is done in the <prgn>postinst</prgn>, it
            should be protected with a conditional so that
            unnecessary prompting doesn't happen if a package's
            installation fails and the <prgn>postinst</prgn> is
            called with <tt>abort-upgrade</tt>,
            <tt>abort-remove</tt> or <tt>abort-deconfigure</tt>.
          </p>
        </sect1>

      </sect>

    </chapt>


    <chapt id="source">
      <heading>Source packages</heading>

      <sect id="standardsversion">
        <heading>Standards conformance</heading>

        <p>
          Source packages should specify the most recent version number
          of this policy document with which your package complied
          when it was last updated.
        </p>

        <p>
          This information may be used to file bug reports
          automatically if your package becomes too much out of date.
        </p>

        <p>
          The version is specified in the <tt>Standards-Version</tt>
          control field.
          The format of the <tt>Standards-Version</tt> field is
          described in <ref id="f-Standards-Version">.
        </p>

        <p>
          You should regularly, and especially if your package has
          become out of date, check for the newest Policy Manual
          available and update your package, if necessary. When your
          package complies with the new standards you should update the
          <tt>Standards-Version</tt> source package field and
          release it.<footnote>
                See the file <file>upgrading-checklist</file> for
                information about policy which has changed between
                different versions of this document.
          </footnote>
        </p>

      </sect>

      <sect id="pkg-relations">
        <heading>Package relationships</heading>

        <p>
          Source packages should specify which binary packages they
          require to be installed or not to be installed in order to
          build correctly.  For example, if building a package
          requires a certain compiler, then the compiler should be
          specified as a build-time dependency.
        </p>

        <p>
          It is not necessary to explicitly specify build-time
          relationships on a minimal set of packages that are always
          needed to compile, link and put in a Debian package a
          standard "Hello World!" program written in C or C++.  The
          required packages are called <em>build-essential</em>, and
          an informational list can be found in
          <file>/usr/share/doc/build-essential/list</file> (which is
          contained in the <tt>build-essential</tt>
          package).<footnote>
            Rationale:
                <list compact="compact">
                  <item>
                      This allows maintaining the list separately
                      from the policy documents (the list does not
                      need the kind of control that the policy
                      documents do).
                  </item>
                  <item>
                      Having a separate package allows one to install
                      the build-essential packages on a machine, as
                      well as allowing other packages such as tasks to
                      require installation of the build-essential
                      packages using the depends relation.
                  </item>
                  <item>
                      The separate package allows bug reports against
                      the list to be categorized separately from
                      the policy management process in the BTS.
                  </item>
                </list>
          </footnote>
        </p>

        <p>
          When specifying the set of build-time dependencies, one
          should list only those packages explicitly required by the
          build.  It is not necessary to list packages which are
          required merely because some other package in the list of
          build-time dependencies depends on them.<footnote>
                The reason for this is that dependencies change, and
                you should list all those packages, and <em>only</em>
                those packages that <em>you</em> need directly.  What
                others need is their business.  For example, if you
                only link against <file>libimlib</file>, you will need to
                build-depend on <package>libimlib2-dev</package> but
                not against any <tt>libjpeg*</tt> packages, even
                though <tt>libimlib2-dev</tt> currently depends on
                them: installation of <package>libimlib2-dev</package>
                will automatically ensure that all of its run-time
                dependencies are satisfied.
          </footnote>
        </p>

        <p>
          If build-time dependencies are specified, it must be
          possible to build the package and produce working binaries
          on a system with only essential and build-essential
          packages installed and also those required to satisfy the
          build-time relationships (including any implied
          relationships).  In particular, this means that version
          clauses should be used rigorously in build-time
          relationships so that one cannot produce bad or
          inconsistently configured packages when the relationships
          are properly satisfied.
        </p>

        <p>
          <ref id="relationships"> explains the technical details.
        </p>
      </sect>

      <sect>
        <heading>Changes to the upstream sources</heading>

        <p>
          If changes to the source code are made that are not
          specific to the needs of the Debian system, they should be
          sent to the upstream authors in whatever form they prefer
          so as to be included in the upstream version of the
          package.
        </p>

        <p>
          If you need to configure the package differently for
          Debian or for Linux, and the upstream source doesn't
          provide a way to do so, you should add such configuration
          facilities (for example, a new <prgn>autoconf</prgn> test
          or <tt>#define</tt>) and send the patch to the upstream
          authors, with the default set to the way they originally
          had it.  You can then easily override the default in your
          <file>debian/rules</file> or wherever is appropriate.
        </p>

        <p>
          You should make sure that the <prgn>configure</prgn> utility
          detects the correct architecture specification string
          (refer to <ref id="arch-spec"> for details).
        </p>

        <p>
          If you need to edit a <prgn>Makefile</prgn> where GNU-style
          <prgn>configure</prgn> scripts are used, you should edit the
          <file>.in</file> files rather than editing the
          <prgn>Makefile</prgn> directly.  This allows the user to
          reconfigure the package if necessary.  You should
          <em>not</em> configure the package and edit the generated
          <prgn>Makefile</prgn>!  This makes it impossible for someone
          else to later reconfigure the package without losing the
          changes you made.
        </p>

      </sect>

      <sect id="dpkgchangelog">
        <heading>Debian changelog: <file>debian/changelog</file></heading>

        <p>
          Changes in the Debian version of the package should be
          briefly explained in the Debian changelog file
          <file>debian/changelog</file>.<footnote>
            <p>
              Mistakes in changelogs are usually best rectified by
              making a new changelog entry rather than "rewriting
              history" by editing old changelog entries.
            </p>
          </footnote>
          This includes modifications
          made in the Debian package compared to the upstream one
          as well as other changes and updates to the package.
          <footnote>
              Although there is nothing stopping an author who is also
              the Debian maintainer from using this changelog for all
              their changes, it will have to be renamed if the Debian
              and upstream maintainers become different people. In such
              a case, however, it might be better to maintain the package
              as a non-native package.
          </footnote>
        </p>

        <p>
          The format of the <file>debian/changelog</file> allows the
          package building tools to discover which version of the package
          is being built and find out other release-specific information.
        </p>

        <p>
          That format is a series of entries like this:

<example compact="compact">
<var>package</var> (<var>version</var>) <var>distribution(s)</var>; urgency=<var>urgency</var>
            <var>
              [optional blank line(s), stripped]
            </var>
  * <var>change details</var>
    <var>more change details</var>
            <var>
              [blank line(s), included in output of dpkg-parsechangelog]
            </var>
  * <var>even more change details</var>
            <var>
              [optional blank line(s), stripped]
            </var>
 -- <var>maintainer name</var> &lt;<var>email address</var>&gt;<var>[two spaces]</var>  <var>date</var>
</example>
        </p>

        <p>
          <var>package</var> and <var>version</var> are the source
          package name and version number.
        </p>

        <p>
          <var>distribution(s)</var> lists the distributions where
          this version should be installed when it is uploaded - it
          is copied to the <tt>Distribution</tt> field in the
          <file>.changes</file> file.  See <ref id="f-Distribution">.
        </p>

        <p>
          <var>urgency</var> is the value for the <tt>Urgency</tt>
          field in the <file>.changes</file> file for the upload
          (see <ref id="f-Urgency">). It is not possible to specify
          an urgency containing commas; commas are used to separate
          <tt><var>keyword</var>=<var>value</var></tt> settings in the
          <prgn>dpkg</prgn> changelog format (though there is
          currently only one useful <var>keyword</var>,
          <tt>urgency</tt>).
        </p>

        <p>
          The change details may in fact be any series of lines
          starting with at least two spaces, but conventionally each
          change starts with an asterisk and a separating space and
          continuation lines are indented so as to bring them in
          line with the start of the text above.  Blank lines may be
          used here to separate groups of changes, if desired.
        </p>

        <p>
          If this upload resolves bugs recorded in the Bug Tracking
          System (BTS), they may be automatically closed on the
          inclusion of this package into the Debian archive by
          including the string: <tt>closes: Bug#<var>nnnnn</var></tt>
          in the change details.<footnote>
              To be precise, the string should match the following
              Perl regular expression:
              <example>
/closes:\s*(?:bug)?\#?\s?\d+(?:,\s*(?:bug)?\#?\s?\d+)*/i
              </example>
              Then all of the bug numbers listed will be closed by the
              archive maintenance software (<prgn>dak</prgn>) using the
              <var>version</var> of the changelog entry.
          </footnote>
          This information is conveyed via the <tt>Closes</tt> field
          in the <tt>.changes</tt> file (see <ref id="f-Closes">).
        </p>

        <p>
          The maintainer name and email address used in the changelog
          should be the details of the person who prepared this release of
          the package.  They are <em>not</em> necessarily those of the
          uploader or usual package maintainer.<footnote>
            In the case of a sponsored upload, the uploader signs the
            files, but the changelog maintainer name and address are those
            of the person who prepared this release.  If the preparer of
            the release is not one of the usual maintainers of the package
            (as listed in
            the <qref id="f-Maintainer"><tt>Maintainer</tt></qref>
            or <qref id="f-Uploaders"><tt>Uploaders</tt></qref> control
            fields of the package), the first line of the changelog is
            conventionally used to explain why a non-maintainer is
            uploading the package.  The Debian Developer's Reference
            (see <ref id="related">) documents the conventions
            used.</footnote>
          The information here will be copied to the <tt>Changed-By</tt>
          field in the <tt>.changes</tt> file
          (see <ref id="f-Changed-By">), and then later used to send an
          acknowledgement when the upload has been installed.
        </p>

        <p>
          The <var>date</var> has the following format<footnote>
              This is the same as the format generated by <tt>date
              -R</tt>.
          </footnote> (compatible and with the same semantics of
          RFC 2822 and RFC 5322):
          <example>day-of-week, dd month yyyy hh:mm:ss +zzzz</example>
          where:
          <list compact="compact">
            <item>
              day-of week is one of: Mon, Tue, Wed, Thu, Fri, Sat, Sun
            </item>
            <item>
              dd is a one- or two-digit day of the month (01-31)
            </item>
            <item>
              month is one of: Jan, Feb, Mar, Apr, May, Jun, Jul, Aug,
              Sep, Oct, Nov, Dec
            </item>
            <item>yyyy is the four-digit year (e.g. 2010)</item>
            <item>hh is the two-digit hour (00-23)</item>
            <item>mm is the two-digit minutes (00-59)</item>
            <item>ss is the two-digit seconds (00-60)</item>
            <item>
              +zzzz or -zzzz is the the time zone offset from Coordinated
              Universal Time (UTC).  "+" indicates that the time is ahead
              of (i.e., east of) UTC and "-" indicates that the time is
              behind (i.e., west of) UTC.  The first two digits indicate
              the hour difference from UTC and the last two digits
              indicate the number of additional minutes difference from
              UTC.  The last two digits must be in the range 00-59.
            </item>
          </list>
        </p>

        <p>
          The first "title" line with the package name must start
          at the left hand margin.  The "trailer" line with the
          maintainer and date details must be preceded by exactly
          one space.  The maintainer details and the date must be
          separated by exactly two spaces.
        </p>

        <p>
          The entire changelog must be encoded in UTF-8.
        </p>

        <p>
          For more information on placement of the changelog files
          within binary packages, please see <ref id="changelogs">.
        </p>
      </sect>

      <sect id="dpkgcopyright">
        <heading>Copyright: <file>debian/copyright</file></heading>
        <p>
          Every package must be accompanied by a verbatim copy of its
          copyright information and distribution license in the file
          <file>/usr/share/doc/<var>package</var>/copyright</file>
          (see <ref id="copyrightfile"> for further details). Also see
          <ref id="pkgcopyright"> for further considerations related
          to copyrights for packages.
        </p>
      </sect>
      <sect>
        <heading>Error trapping in makefiles</heading>

        <p>
          When <prgn>make</prgn> invokes a command in a makefile
          (including your package's upstream makefiles and
          <file>debian/rules</file>), it does so using <prgn>sh</prgn>.  This
          means that <prgn>sh</prgn>'s usual bad error handling
          properties apply: if you include a miniature script as one
          of the commands in your makefile you'll find that if you
          don't do anything about it then errors are not detected
          and <prgn>make</prgn> will blithely continue after
          problems.
        </p>

        <p>
          Every time you put more than one shell command (this
          includes using a loop) in a makefile command you
          must make sure that errors are trapped.  For
          simple compound commands, such as changing directory and
          then running a program, using <tt>&amp;&amp;</tt> rather
          than semicolon as a command separator is sufficient.  For
          more complex commands including most loops and
          conditionals you should include a separate <tt>set -e</tt>
          command at the start of every makefile command that's
          actually one of these miniature shell scripts.
        </p>
      </sect>

      <sect id="timestamps">
        <heading>Time Stamps</heading>
        <p>
          Maintainers should preserve the modification times of the
          upstream source files in a package, as far as is reasonably
          possible.<footnote>
              The rationale is that there is some information conveyed
              by knowing the age of the file, for example, you could
              recognize that some documentation is very old by looking
              at the modification time, so it would be nice if the
              modification time of the upstream source would be
              preserved.
          </footnote>
        </p>
      </sect>

      <sect id="restrictions">
        <heading>Restrictions on objects in source packages</heading>

        <p>
          The source package may not contain any hard links<footnote>
            <p>
              This is not currently detected when building source
              packages, but only when extracting
              them.
            </p>
            <p>
              Hard links may be permitted at some point in the
              future, but would require a fair amount of
              work.
            </p>
          </footnote>, device special files, sockets or setuid or
          setgid files.<footnote>
              Setgid directories are allowed.
          </footnote>
        </p>
      </sect>

      <sect id="debianrules">
        <heading>Main building script: <file>debian/rules</file></heading>

        <p>
          This file must be an executable makefile, and contains the
          package-specific recipes for compiling the package and
          building binary package(s) from the source.
        </p>

        <p>
          It must start with the line <tt>#!/usr/bin/make -f</tt>,
          so that it can be invoked by saying its name rather than
          invoking <prgn>make</prgn> explicitly. That is, invoking
          either of <tt>make -f debian/rules <em>args...</em></tt>
          or <tt>./debian/rules <em>args...</em></tt> must result in
          identical behavior.
        </p>

        <p>
          The following targets are required and must be implemented
          by <file>debian/rules</file>: <tt>clean</tt>, <tt>binary</tt>,
          <tt>binary-arch</tt>, <tt>binary-indep</tt>, <tt>build</tt>,
          <tt>build-arch</tt> and <tt>build-indep</tt>.
          These are the targets called by <prgn>dpkg-buildpackage</prgn>.
        </p>

        <p>
          Since an interactive <file>debian/rules</file> script makes it
          impossible to auto-compile that package and also makes it hard
          for other people to reproduce the same binary package, all
          required targets must be non-interactive.  It also follows that
          any target that these targets depend on must also be
          non-interactive.
        </p>

        <p>
          The targets are as follows:
          <taglist>
            <tag><tt>build</tt> (required)</tag>
            <item>
              <p>
                The <tt>build</tt> target should perform all the
                configuration and compilation of the package.
                If a package has an interactive pre-build
                configuration routine, the Debian source package
                must either be built after this has taken place (so
                that the binary package can be built without rerunning
                the configuration) or the configuration routine
                modified to become non-interactive.  (The latter is
                preferable if there are architecture-specific features
                detected by the configuration routine.)
              </p>

              <p>
                For some packages, notably ones where the same
                source tree is compiled in different ways to produce
                two binary packages, the <tt>build</tt> target
                does not make much sense.  For these packages it is
                good enough to provide two (or more) targets
                (<tt>build-a</tt> and <tt>build-b</tt> or whatever)
                for each of the ways of building the package, and a
                <tt>build</tt> target that does nothing.  The
                <tt>binary</tt> target will have to build the
                package in each of the possible ways and make the
                binary package out of each.
              </p>

              <p>
                The <tt>build</tt> target must not do anything
                that might require root privilege.
              </p>

              <p>
                The <tt>build</tt> target may need to run the
                <tt>clean</tt> target first - see below.
              </p>

              <p>
                When a package has a configuration and build routine
                which takes a long time, or when the makefiles are
                poorly designed, or when <tt>build</tt> needs to
                run <tt>clean</tt> first, it is a good idea to
                <tt>touch build</tt> when the build process is
                complete.  This will ensure that if <tt>debian/rules
                build</tt> is run again it will not rebuild the whole
                program.<footnote>
                    Another common way to do this is for <tt>build</tt>
                    to depend on <prgn>build-stamp</prgn> and to do
                    nothing else, and for the <prgn>build-stamp</prgn>
                    target to do the building and to <tt>touch
                    build-stamp</tt> on completion.  This is
                    especially useful if the build routine creates a
                    file or directory called <tt>build</tt>; in such a
                    case, <tt>build</tt> will need to be listed as
                    a phony target (i.e., as a dependency of the
                    <tt>.PHONY</tt> target).  See the documentation of
                    <prgn>make</prgn> for more information on phony
                    targets.
                </footnote>
              </p>
            </item>

            <tag><tt>build-arch</tt> (required),
                 <tt>build-indep</tt> (required)
            </tag>
            <item>
              <p>
                The <tt>build-arch</tt> target must
                perform all the configuration and compilation required for
                producing all architecture-dependant binary packages
                (those packages for which the body of the
                <tt>Architecture</tt> field in <tt>debian/control</tt> is
                not <tt>all</tt>).  Similarly, the <tt>build-indep</tt>
                target must perform all the configuration
                and compilation required for producing all
                architecture-independent binary packages (those packages
                for which the body of the <tt>Architecture</tt> field
                in <tt>debian/control</tt> is <tt>all</tt>).
                The <tt>build</tt> target
                should either depend on those targets or take the same
                actions as invoking those targets would perform.<footnote>
                  This split allows binary-only builds to not install the
                  dependencies required for the <tt>build-indep</tt>
                  target and skip any resource-intensive build tasks that
                  are only required when building architecture-independent
                  binary packages.
                </footnote>
              </p>

              <p>
                The <tt>build-arch</tt> and <tt>build-indep</tt> targets
                must not do anything that might require root privilege.
              </p>
            </item>

            <tag><tt>binary</tt> (required), <tt>binary-arch</tt>
              (required), <tt>binary-indep</tt> (required)
            </tag>
            <item>
              <p>
                The <tt>binary</tt> target must be all that is
                necessary for the user to build the binary package(s)
                produced from this source package.  It is
                split into two parts: <prgn>binary-arch</prgn> builds
                the binary packages which are specific to a particular
                architecture, and <tt>binary-indep</tt> builds
                those which are not.
              </p>
              <p>
                <tt>binary</tt> may be (and commonly is) a target with
                no commands which simply depends on
                <tt>binary-arch</tt> and <tt>binary-indep</tt>.
              </p>
              <p>
                Both <tt>binary-*</tt> targets should depend on the
                <tt>build</tt> target, or on the appropriate
                <tt>build-arch</tt> or <tt>build-indep</tt> target, if
                provided, so that the package is built if it has not
                been already.  It should then create the relevant
                binary package(s), using <prgn>dpkg-gencontrol</prgn> to
                make their control files and <prgn>dpkg-deb</prgn> to
                build them and place them in the parent of the top
                level directory.
              </p>

              <p>
                Both the <tt>binary-arch</tt> and
                <tt>binary-indep</tt> targets <em>must</em> exist.
                If one of them has nothing to do (which will always be
                the case if the source generates only a single binary
                package, whether architecture-dependent or not), it
                must still exist and must always succeed.
              </p>

              <p>
                The <tt>binary</tt> targets must be invoked as
                root.<footnote>
                    The <prgn>fakeroot</prgn> package often allows one
                    to build a package correctly even without being
                    root.
                </footnote>
              </p>
            </item>

            <tag><tt>clean</tt> (required)</tag>
            <item>
              <p>
                This must undo any effects that the <tt>build</tt>
                and <tt>binary</tt> targets may have had, except
                that it should leave alone any output files created in
                the parent directory by a run of a <tt>binary</tt>
                target.
              </p>

              <p>
                If a <tt>build</tt> file is touched at the end of
                the <tt>build</tt> target, as suggested above, it
                should be removed as the first action that
                <tt>clean</tt> performs, so that running
                <tt>build</tt> again after an interrupted
                <tt>clean</tt> doesn't think that everything is
                already done.
              </p>

              <p>
                The <tt>clean</tt> target may need to be
                invoked as root if <tt>binary</tt> has been
                invoked since the last <tt>clean</tt>, or if
                <tt>build</tt> has been invoked as root (since
                <tt>build</tt> may create directories, for
                example).
              </p>
            </item>

            <tag><tt>get-orig-source</tt> (optional)</tag>
            <item>
              <p>
                This target fetches the most recent version of the
                original source package from a canonical archive site
                (via FTP or WWW, for example), does any necessary
                rearrangement to turn it into the original source
                tar file format described below, and leaves it in the
                current directory.
              </p>

              <p>
                This target may be invoked in any directory, and
                should take care to clean up any temporary files it
                may have left.
              </p>

              <p>
                This target is optional, but providing it if
                possible is a good idea.
              </p>
            </item>

            <tag><tt>patch</tt> (optional)</tag>
            <item>
              <p>
                This target performs whatever additional actions are
                required to make the source ready for editing (unpacking
                additional upstream archives, applying patches, etc.).
                It is recommended to be implemented for any package where
                <tt>dpkg-source -x</tt> does not result in source ready
                for additional modification.  See
                <ref id="readmesource">.
              </p>
            </item>
          </taglist>

        <p>
          The <tt>build</tt>, <tt>binary</tt> and
          <tt>clean</tt> targets must be invoked with the current
          directory being the package's top-level directory.
        </p>


        <p>
          Additional targets may exist in <file>debian/rules</file>,
          either as published or undocumented interfaces or for the
          package's internal use.
        </p>

        <p>
          The architectures we build on and build for are determined
          by <prgn>make</prgn> variables using the
          utility <prgn>dpkg-architecture</prgn>.
          You can determine the Debian architecture and the GNU style
          architecture specification string for the build architecture as
          well as for the host architecture.  The build architecture is
          the architecture on which <file>debian/rules</file> is run and
          the package build is performed.  The host architecture is the
          architecture on which the resulting package will be installed
          and run.  These are normally the same, but may be different in
          the case of cross-compilation (building packages for one
          architecture on machines of a different architecture).
        </p>

        <p>
          Here is a list of supported <prgn>make</prgn> variables:
          <list compact="compact">
            <item>
                <tt>DEB_*_ARCH</tt> (the Debian architecture)
            </item>
            <item>
                <tt>DEB_*_ARCH_CPU</tt> (the Debian CPU name)
            </item>
            <item>
                <tt>DEB_*_ARCH_OS</tt> (the Debian System name)
            </item>
            <item>
                <tt>DEB_*_GNU_TYPE</tt> (the GNU style architecture
                specification string)
            </item>
            <item>
                <tt>DEB_*_GNU_CPU</tt> (the CPU part of
                <tt>DEB_*_GNU_TYPE</tt>)
            </item>
            <item>
                <tt>DEB_*_GNU_SYSTEM</tt> (the System part of
                <tt>DEB_*_GNU_TYPE</tt>)
          </list>
          where <tt>*</tt> is either <tt>BUILD</tt> for specification of
          the build architecture or <tt>HOST</tt> for specification of the
          host architecture.
        </p>

        <p>
          Backward compatibility can be provided in the rules file
          by setting the needed variables to suitable default
          values; please refer to the documentation of
          <prgn>dpkg-architecture</prgn> for details.
        </p>

        <p>
          It is important to understand that the <tt>DEB_*_ARCH</tt>
          string only determines which Debian architecture we are
          building on or for. It should not be used to get the CPU
          or system information; the <tt>DEB_*_ARCH_CPU</tt> and
          <tt>DEB_*_ARCH_OS</tt> variables should be used for that.
          GNU style variables should generally only be used with upstream
          build systems.
        </p>

        <sect1 id="debianrules-options">
          <heading><file>debian/rules</file> and
            <tt>DEB_BUILD_OPTIONS</tt></heading>

          <p>
            Supporting the standardized environment variable
            <tt>DEB_BUILD_OPTIONS</tt> is recommended.  This variable can
            contain several flags to change how a package is compiled and
            built.  Each flag must be in the form <var>flag</var> or
            <var>flag</var>=<var>options</var>.  If multiple flags are
            given, they must be separated by whitespace.<footnote>
              Some packages support any delimiter, but whitespace is the
              easiest to parse inside a makefile and avoids ambiguity with
              flag values that contain commas.
            </footnote>
            <var>flag</var> must start with a lowercase letter
            (<tt>a-z</tt>) and consist only of lowercase letters,
            numbers (<tt>0-9</tt>), and the characters
            <tt>-</tt> and <tt>_</tt> (hyphen and underscore).
            <var>options</var> must not contain whitespace.  The same
            tag should not be given multiple times with conflicting
            values.  Package maintainers may assume that
            <tt>DEB_BUILD_OPTIONS</tt> will not contain conflicting tags.
          </p>

          <p>
            The meaning of the following tags has been standardized:
            <taglist>
              <tag>nocheck</tag>
              <item>
                  This tag says to not run any build-time test suite
                  provided by the package.
              </item>
              <tag>noopt</tag>
              <item>
                  The presence of this tag means that the package should
                  be compiled with a minimum of optimization.  For C
                  programs, it is best to add <tt>-O0</tt> to
                  <tt>CFLAGS</tt> (although this is usually the default).
                  Some programs might fail to build or run at this level
                  of optimization; it may be necessary to use
                  <tt>-O1</tt>, for example.
              </item>
              <tag>nostrip</tag>
              <item>
                  This tag means that the debugging symbols should not be
                  stripped from the binary during installation, so that
                  debugging information may be included in the package.
              </item>
              <tag>parallel=n</tag>
              <item>
                  This tag means that the package should be built using up
                  to <tt>n</tt> parallel processes if the package build
                  system supports this.<footnote>
                      Packages built with <tt>make</tt> can often implement
                      this by passing the <tt>-j</tt><var>n</var> option to
                      <tt>make</tt>.
                  </footnote>
                  If the package build system does not support parallel
                  builds, this string must be ignored.  If the package
                  build system only supports a lower level of concurrency
                  than <var>n</var>, the package should be built using as
                  many parallel processes as the package build system
                  supports.  It is up to the package maintainer to decide
                  whether the package build times are long enough and the
                  package build system is robust enough to make supporting
                  parallel builds worthwhile.
               </item>
            </taglist>
          </p>

          <p>
            Unknown flags must be ignored by <file>debian/rules</file>.
          </p>

          <p>
            The following makefile snippet is an example of how one may
            implement the build options; you will probably have to
            massage this example in order to make it work for your
            package.
            <example compact="compact">
CFLAGS = -Wall -g
INSTALL = install
INSTALL_FILE    = $(INSTALL) -p    -o root -g root  -m  644
INSTALL_PROGRAM = $(INSTALL) -p    -o root -g root  -m  755
INSTALL_SCRIPT  = $(INSTALL) -p    -o root -g root  -m  755
INSTALL_DIR     = $(INSTALL) -p -d -o root -g root  -m  755

ifneq (,$(filter noopt,$(DEB_BUILD_OPTIONS)))
    CFLAGS += -O0
else
    CFLAGS += -O2
endif
ifeq (,$(filter nostrip,$(DEB_BUILD_OPTIONS)))
    INSTALL_PROGRAM += -s
endif
ifneq (,$(filter parallel=%,$(DEB_BUILD_OPTIONS)))
    NUMJOBS = $(patsubst parallel=%,%,$(filter parallel=%,$(DEB_BUILD_OPTIONS)))
    MAKEFLAGS += -j$(NUMJOBS)
endif

build:
        # ...
ifeq (,$(filter nocheck,$(DEB_BUILD_OPTIONS)))
        # Code to run the package test suite.
endif
            </example>
          </p>
        </sect1>
      </sect>

<!-- FIXME: section pkg-srcsubstvars is the same as substvars -->
      <sect id="substvars">
        <heading>Variable substitutions: <file>debian/substvars</file></heading>

        <p>
          When <prgn>dpkg-gencontrol</prgn>
          generates <qref id="binarycontrolfiles">binary package control
          files</qref> (<file>DEBIAN/control</file>), it performs variable
          substitutions on its output just before writing it.  Variable
          substitutions have the form <tt>${<var>variable</var>}</tt>.
          The optional file <file>debian/substvars</file> contains
          variable substitutions to be used; variables can also be set
          directly from <file>debian/rules</file> using the <tt>-V</tt>
          option to the source packaging commands, and certain predefined
          variables are also available.
        </p>

        <p>
          The <file>debian/substvars</file> file is usually generated and
          modified dynamically by <file>debian/rules</file> targets, in
          which case it must be removed by the <tt>clean</tt> target.
        </p>

        <p>
          See <manref name="deb-substvars" section="5"> for full
          details about source variable substitutions, including the
          format of <file>debian/substvars</file>.</p>
      </sect>

      <sect id="debianwatch">
        <heading>Optional upstream source location: <file>debian/watch</file></heading>

        <p>
          This is an optional, recommended configuration file for the
          <tt>uscan</tt> utility which defines how to automatically scan
          ftp or http sites for newly available updates of the
          package. This is used Debian QA
          tools to help with quality control and maintenance of the
          distribution as a whole.
        </p>

      </sect>

      <sect id="debianfiles">
        <heading>Generated files list: <file>debian/files</file></heading>

        <p>
          This file is not a permanent part of the source tree; it
          is used while building packages to record which files are
          being generated.  <prgn>dpkg-genchanges</prgn> uses it
          when it generates a <file>.changes</file> file.
        </p>

        <p>
          It should not exist in a shipped source package, and so it
          (and any backup files or temporary files such as
          <file>files.new</file><footnote>
              <file>files.new</file> is used as a temporary file by
              <prgn>dpkg-gencontrol</prgn> and
              <prgn>dpkg-distaddfile</prgn> - they write a new
              version of <tt>files</tt> here before renaming it,
              to avoid leaving a corrupted copy if an error
              occurs.
          </footnote>) should be removed by the
          <tt>clean</tt> target.  It may also be wise to
          ensure a fresh start by emptying or removing it at the
          start of the <tt>binary</tt> target.
        </p>

        <p>
          When <prgn>dpkg-gencontrol</prgn> is run for a binary
          package, it adds an entry to <file>debian/files</file> for the
          <file>.deb</file> file that will be created when <tt>dpkg-deb
          --build</tt> is run for that binary package.  So for most
          packages all that needs to be done with this file is to
          delete it in the <tt>clean</tt> target.
        </p>

        <p>
          If a package upload includes files besides the source
          package and any binary packages whose control files were
          made with <prgn>dpkg-gencontrol</prgn> then they should be
          placed in the parent of the package's top-level directory
          and <prgn>dpkg-distaddfile</prgn> should be called to add
          the file to the list in <file>debian/files</file>.</p>
      </sect>

      <sect id="embeddedfiles">
        <heading>Convenience copies of code</heading>

        <p>
          Some software packages include in their distribution convenience
          copies of code from other software packages, generally so that
          users compiling from source don't have to download multiple
          packages.  Debian packages should not make use of these
          convenience copies unless the included package is explicitly
          intended to be used in this way.<footnote>
            For example, parts of the GNU build system work like this.
          </footnote>
          If the included code is already in the Debian archive in the
          form of a library, the Debian packaging should ensure that
          binary packages reference the libraries already in Debian and
          the convenience copy is not used.  If the included code is not
          already in Debian, it should be packaged separately as a
          prerequisite if possible.
          <footnote>
            Having multiple copies of the same code in Debian is
            inefficient, often creates either static linking or shared
            library conflicts, and, most importantly, increases the
            difficulty of handling security vulnerabilities in the
            duplicated code.
          </footnote>
        </p>
      </sect>

      <sect id="readmesource">
        <heading>Source package handling:
          <file>debian/README.source</file></heading>

        <p>
          If running <prgn>dpkg-source -x</prgn> on a source package
          doesn't produce the source of the package, ready for editing,
          and allow one to make changes and run
          <prgn>dpkg-buildpackage</prgn> to produce a modified package
          without taking any additional steps, creating a
          <file>debian/README.source</file> documentation file is
          recommended.  This file should explain how to do all of the
          following:
            <enumlist>
              <item>Generate the fully patched source, in a form ready for
              editing, that would be built to create Debian
              packages.  Doing this with a <tt>patch</tt> target in
              <file>debian/rules</file> is recommended; see
              <ref id="debianrules">.</item>
              <item>Modify the source and save those modifications so that
              they will be applied when building the package.</item>
              <item>Remove source modifications that are currently being
              applied when building the package.</item>
              <item>Optionally, document what steps are necessary to
              upgrade the Debian source package to a new upstream version,
              if applicable.</item>
            </enumlist>
          This explanation should include specific commands and mention
          any additional required Debian packages.  It should not assume
          familiarity with any specific Debian packaging system or patch
          management tools.
        </p>

        <p>
          This explanation may refer to a documentation file installed by
          one of the package's build dependencies provided that the
          referenced documentation clearly explains these tasks and is not
          a general reference manual.
        </p>

        <p>
          <file>debian/README.source</file> may also include any other
          information that would be helpful to someone modifying the
          source package.  Even if the package doesn't fit the above
          description, maintainers are encouraged to document in a
          <file>debian/README.source</file> file any source package with a
          particularly complex or unintuitive source layout or build
          system (for example, a package that builds the same source
          multiple times to generate different binary packages).
        </p>
      </sect>
    </chapt>


    <chapt id="controlfields">
      <heading>Control files and their fields</heading>

      <p>
        The package management system manipulates data represented in
        a common format, known as <em>control data</em>, stored in
        <em>control files</em>.
        Control files are used for source packages, binary packages and
        the <file>.changes</file> files which control the installation
        of uploaded files<footnote>
            <prgn>dpkg</prgn>'s internal databases are in a similar
            format.
        </footnote>.
      </p>

      <sect id="controlsyntax">
        <heading>Syntax of control files</heading>

        <p>
          A control file consists of one or more paragraphs of
          fields<footnote>
                The paragraphs are also sometimes referred to as stanzas.
          </footnote>.
          The paragraphs are separated by empty lines.  Parsers may accept
          lines consisting solely of spaces and tabs as paragraph
          separators, but control files should use empty lines.  Some control
          files allow only one paragraph; others allow several, in
          which case each paragraph usually refers to a different
          package.  (For example, in source packages, the first
          paragraph refers to the source package, and later paragraphs
          refer to binary packages generated from the source.)  The
          ordering of the paragraphs in control files is significant.
        </p>

        <p>
          Each paragraph consists of a series of data fields.  Each field
          consists of the field name followed by a colon and then the
          data/value associated with that field.  The field name is
          composed of US-ASCII characters excluding control characters,
          space, and colon (i.e., characters in the ranges 33-57 and
          59-126, inclusive).  Field names must not begin with the comment
          character, <tt>#</tt>, nor with the hyphen character, <tt>-</tt>.
        </p>

        <p>
          The field ends at the end of the line or at the end of the last
          continuation line (see below).  Horizontal whitespace (spaces
          and tabs) may occur immediately before or after the value and is
          ignored there; it is conventional to put a single space after
          the colon.  For example, a field might be:
          <example compact="compact">
Package: libc6
          </example>
          the field name is <tt>Package</tt> and the field value
          <tt>libc6</tt>.
        </p>

        <p>
          A paragraph must not contain more than one instance of a
          particular field name.
        </p>

        <p>
          There are three types of fields:
          <taglist>
            <tag>simple</tag>
            <item>
              The field, including its value, must be a single line.  Folding
              of the field is not permitted.  This is the default field type
              if the definition of the field does not specify a different
              type.
            </item>
            <tag>folded</tag>
            <item>
              The value of a folded field is a logical line that may span
              several lines.  The lines after the first are called
              continuation lines and must start with a space or a tab.
              Whitespace, including any newlines, is not significant in the
              field values of folded fields.<footnote>
                This folding method is similar to RFC 5322, allowing control
                files that contain only one paragraph and no multiline fields
                to be read by parsers written for RFC 5322.
              </footnote>
            </item>
            <tag>multiline</tag>
            <item>
              The value of a multiline field may comprise multiple continuation
              lines.  The first line of the value, the part on the same line as
              the field name, often has special significance or may have to be
              empty.  Other lines are added following the same syntax as the
              continuation lines of the folded fields.  Whitespace, including newlines,
              is significant in the values of multiline fields.
            </item>
          </taglist>
        </p>

        <p>
          Whitespace must not appear
          inside names (of packages, architectures, files or anything
          else) or version numbers, or between the characters of
          multi-character version relationships.
        </p>

        <p>
          The presence and purpose of a field, and the syntax of its
          value may differ between types of control files.
        </p>

        <p>
          Field names are not case-sensitive, but it is usual to
          capitalize the field names using mixed case as shown below.
          Field values are case-sensitive unless the description of the
          field says otherwise.
        </p>

        <p>
          Paragraph separators (empty lines) and lines consisting only of
          spaces and tabs are not allowed within field values or between
          fields.  Empty lines in field values are usually escaped by
          representing them by a space followed by a dot.
        </p>

        <p>
          Lines starting with # without any preceding whitespace are comments
          lines that are only permitted in source package control files
          (<file>debian/control</file>).  These comment lines are ignored, even
          between two continuation lines.  They do not end logical lines.
        </p>

        <p>
          All control files must be encoded in UTF-8.
        </p>
      </sect>

      <sect id="sourcecontrolfiles">
        <heading>Source package control files -- <file>debian/control</file></heading>

        <p>
          The <file>debian/control</file> file contains the most vital
          (and version-independent) information about the source package
          and about the binary packages it creates.
        </p>

        <p>
          The first paragraph of the control file contains information about
          the source package in general. The subsequent sets each describe a
          binary package that the source tree builds.
        </p>

        <p>
          The fields in the general paragraph (the first one, for the source
          package) are:

          <list compact="compact">
            <item><qref id="f-Source"><tt>Source</tt></qref> (mandatory)</item>
            <item><qref id="f-Maintainer"><tt>Maintainer</tt></qref> (mandatory)</item>
            <item><qref id="f-Uploaders"><tt>Uploaders</tt></qref></item>
            <item><qref id="f-Section"><tt>Section</tt></qref> (recommended)</item>
            <item><qref id="f-Priority"><tt>Priority</tt></qref> (recommended)</item>
            <item><qref id="sourcebinarydeps"><tt>Build-Depends</tt> et al</qref></item>
            <item><qref id="f-Standards-Version"><tt>Standards-Version</tt></qref> (recommended)</item>
            <item><qref id="f-Homepage"><tt>Homepage</tt></qref></item>
            <item><qref id="f-VCS-fields"><tt>Vcs-Browser</tt>, <tt>Vcs-Git</tt>, et al.</qref></item>
          </list>
        </p>

        <p>
          The fields in the binary package paragraphs are:

          <list compact="compact">
            <item><qref id="f-Package"><tt>Package</tt></qref> (mandatory)</item>
            <item><qref id="f-Architecture"><tt>Architecture</tt></qref> (mandatory)</item>
            <item><qref id="f-Section"><tt>Section</tt></qref> (recommended)</item>
            <item><qref id="f-Priority"><tt>Priority</tt></qref> (recommended)</item>
            <item><qref id="f-Essential"><tt>Essential</tt></qref></item>
            <item><qref id="binarydeps"><tt>Depends</tt> et al</qref></item>
            <item><qref id="f-Description"><tt>Description</tt></qref> (mandatory)</item>
            <item><qref id="f-Homepage"><tt>Homepage</tt></qref></item>
            <item><qref id="built-using"><tt>Built-Using</tt></qref></item>
            <item><qref id="f-Package-Type"><tt>Package-Type</tt></qref></item>
          </list>
        </p>

        <p>
          The syntax and semantics of the fields are described below.
        </p>

        <p>
          These fields are used by <prgn>dpkg-gencontrol</prgn> to
          generate control files for binary packages (see below), by
          <prgn>dpkg-genchanges</prgn> to generate the
          <file>.changes</file> file to accompany the upload, and by
          <prgn>dpkg-source</prgn> when it creates the
          <file>.dsc</file> source control file as part of a source
          archive.  Some fields are folded in <file>debian/control</file>,
          but not in any other control
          file. These tools are responsible for removing the line
          breaks from such fields when using fields from
          <file>debian/control</file> to generate other control files.
        </p>

        <p>
          The fields here may contain variable references - their
          values will be substituted by <prgn>dpkg-gencontrol</prgn>,
          <prgn>dpkg-genchanges</prgn> or <prgn>dpkg-source</prgn>
          when they generate output control files.
          See <ref id="substvars"> for details.
        </p>
      </sect>

      <sect id="binarycontrolfiles">
        <heading>Binary package control files -- <file>DEBIAN/control</file></heading>

        <p>
          The <file>DEBIAN/control</file> file contains the most vital
          (and version-dependent) information about a binary package.  It
          consists of a single paragraph.
        </p>

        <p>
          The fields in this file are:

          <list compact="compact">
            <item><qref id="f-Package"><tt>Package</tt></qref> (mandatory)</item>
            <item><qref id="f-Source"><tt>Source</tt></qref></item>
            <item><qref id="f-Version"><tt>Version</tt></qref> (mandatory)</item>
            <item><qref id="f-Section"><tt>Section</tt></qref> (recommended)</item>
            <item><qref id="f-Priority"><tt>Priority</tt></qref> (recommended)</item>
            <item><qref id="f-Architecture"><tt>Architecture</tt></qref> (mandatory)</item>
            <item><qref id="f-Essential"><tt>Essential</tt></qref></item>
            <item><qref id="binarydeps"><tt>Depends</tt> et al</qref></item>
            <item><qref id="f-Installed-Size"><tt>Installed-Size</tt></qref></item>
            <item><qref id="f-Maintainer"><tt>Maintainer</tt></qref> (mandatory)</item>
            <item><qref id="f-Description"><tt>Description</tt></qref> (mandatory)</item>
            <item><qref id="f-Homepage"><tt>Homepage</tt></qref></item>
            <item><qref id="built-using"><tt>Built-Using</tt></qref></item>
          </list>
        </p>
      </sect>

      <sect id="debiansourcecontrolfiles">
        <heading>Debian source control files -- <tt>.dsc</tt></heading>

        <p>
          This file consists of a single paragraph, possibly surrounded by
          a PGP signature. The fields of that paragraph are listed below.
          Their syntax is described above, in <ref id="controlsyntax">.

        <list compact="compact">
          <item><qref id="f-Format"><tt>Format</tt></qref> (mandatory)</item>
          <item><qref id="f-Source"><tt>Source</tt></qref> (mandatory)</item>
          <item><qref id="f-Binary"><tt>Binary</tt></qref></item>
          <item><qref id="f-Architecture"><tt>Architecture</tt></qref></item>
          <item><qref id="f-Version"><tt>Version</tt></qref> (mandatory)</item>
          <item><qref id="f-Maintainer"><tt>Maintainer</tt></qref> (mandatory)</item>
          <item><qref id="f-Uploaders"><tt>Uploaders</tt></qref></item>
          <item><qref id="f-Homepage"><tt>Homepage</tt></qref></item>
          <item><qref id="f-VCS-fields"><tt>Vcs-Browser</tt>, <tt>Vcs-Git</tt>, et al.</qref></item>
          <item><qref id="f-Dgit"><tt>Dgit</tt></qref></item>
          <item><qref id="f-Standards-Version"><tt>Standards-Version</tt></qref> (recommended)</item>
          <item><qref id="sourcebinarydeps"><tt>Build-Depends</tt> et al</qref></item>
          <item><qref id="f-Package-List"><tt>Package-List</tt></qref> (recommended)</item>
          <item><qref id="f-Checksums"><tt>Checksums-Sha1</tt>
              and <tt>Checksums-Sha256</tt></qref> (mandatory)</item>
          <item><qref id="f-Files"><tt>Files</tt></qref> (mandatory)</item>
        </list>
        </p>

        <p>
          The Debian source control file is generated by
          <prgn>dpkg-source</prgn> when it builds the source
          archive, from other files in the source package,
          described above.  When unpacking, it is checked against
          the files and directories in the other parts of the
          source package.
        </p>

      </sect>

      <sect id="debianchangesfiles">
        <heading>Debian changes files -- <file>.changes</file></heading>

        <p>
          The <file>.changes</file> files are used by the Debian archive
          maintenance software to process updates to packages. They
          consist of a single paragraph, possibly surrounded by a PGP
          signature. That paragraph contains information from the
          <file>debian/control</file> file and other data about the
          source package gathered via <file>debian/changelog</file>
          and <file>debian/rules</file>.
        </p>

        <p>
          <file>.changes</file> files have a format version that is
          incremented whenever the documented fields or their meaning
          change.  This document describes format &changesversion;.
        </p>

        <p>
          The fields in this file are:

          <list compact="compact">
            <item><qref id="f-Format"><tt>Format</tt></qref> (mandatory)</item>
            <item><qref id="f-Date"><tt>Date</tt></qref> (mandatory)</item>
            <item><qref id="f-Source"><tt>Source</tt></qref> (mandatory)</item>
            <item><qref id="f-Binary"><tt>Binary</tt></qref> (mandatory)</item>
            <item><qref id="f-Architecture"><tt>Architecture</tt></qref> (mandatory)</item>
            <item><qref id="f-Version"><tt>Version</tt></qref> (mandatory)</item>
            <item><qref id="f-Distribution"><tt>Distribution</tt></qref> (mandatory)</item>
            <item><qref id="f-Urgency"><tt>Urgency</tt></qref> (recommended)</item>
            <item><qref id="f-Maintainer"><tt>Maintainer</tt></qref> (mandatory)</item>
            <item><qref id="f-Changed-By"><tt>Changed-By</tt></qref></item>
            <item><qref id="f-Description"><tt>Description</tt></qref> (mandatory)</item>
            <item><qref id="f-Closes"><tt>Closes</tt></qref></item>
            <item><qref id="f-Changes"><tt>Changes</tt></qref> (mandatory)</item>
            <item><qref id="f-Checksums"><tt>Checksums-Sha1</tt>
                and <tt>Checksums-Sha256</tt></qref> (mandatory)</item>
            <item><qref id="f-Files"><tt>Files</tt></qref> (mandatory)</item>
          </list>
        </p>
      </sect>

      <sect id="controlfieldslist">
        <heading>List of fields</heading>

        <sect1 id="f-Source">
          <heading><tt>Source</tt></heading>

          <p>
            This field identifies the source package name.
          </p>

          <p>
            In <file>debian/control</file> or a <file>.dsc</file> file,
            this field must contain only the name of the source package.
          </p>

          <p>
            In a binary package control file or a <file>.changes</file>
            file, the source package name may be followed by a version
            number in parentheses<footnote>
                It is customary to leave a space after the package name
                if a version number is specified.
            </footnote>.
            This version number may be omitted (and is, by
            <prgn>dpkg-gencontrol</prgn>) if it has the same value as
            the <tt>Version</tt> field of the binary package in
            question.  The field itself may be omitted from a binary
            package control file when the source package has the same
            name and version as the binary package.
          </p>

          <p>
            Package names (both source and binary,
            see <ref id="f-Package">) must consist only of lower case
            letters (<tt>a-z</tt>), digits (<tt>0-9</tt>), plus
            (<tt>+</tt>) and minus (<tt>-</tt>) signs, and periods
            (<tt>.</tt>).  They must be at least two characters long and
            must start with an alphanumeric character.
          </p>
        </sect1>

        <sect1 id="f-Maintainer">
          <heading><tt>Maintainer</tt></heading>

          <p>
            The package maintainer's name and email address.  The name
            must come first, then the email address inside angle
            brackets <tt>&lt;&gt;</tt> (in RFC822 format).
          </p>

          <p>
            If the maintainer's name contains a full stop then the
            whole field will not work directly as an email address due
            to a misfeature in the syntax specified in RFC822; a
            program using this field as an address must check for this
            and correct the problem if necessary (for example by
            putting the name in round brackets and moving it to the
            end, and bringing the email address forward).
          </p>

          <p>
            See <ref id="maintainer"> for additional requirements and
            information about package maintainers.
          </p>
        </sect1>

        <sect1 id="f-Uploaders">
          <heading><tt>Uploaders</tt></heading>

          <p>
            List of the names and email addresses of co-maintainers of the
            package, if any. If the package has other maintainers besides
            the one named in the <qref id="f-Maintainer">Maintainer
            field</qref>, their names and email addresses should be listed
            here. The format of each entry is the same as that of the
            Maintainer field, and multiple entries must be comma
            separated.
          </p>

          <p>
            This is normally an optional field, but if
            the <tt>Maintainer</tt> control field names a group of people
            and a shared email address, the <tt>Uploaders</tt> field must
            be present and must contain at least one human with their
            personal email address.
          </p>

          <p>
            The Uploaders field in <file>debian/control</file> can be folded.
          </p>
        </sect1>

        <sect1 id="f-Changed-By">
          <heading><tt>Changed-By</tt></heading>

          <p>
            The name and email address of the person who prepared this
            version of the package, usually a maintainer.  The syntax is
            the same as for the <qref id="f-Maintainer">Maintainer
            field</qref>.
          </p>
        </sect1>

        <sect1 id="f-Section">
          <heading><tt>Section</tt></heading>

          <p>
            This field specifies an application area into which the package
            has been classified. See <ref id="subsections">.
          </p>

          <p>
            When it appears in the <file>debian/control</file> file,
            it gives the value for the subfield of the same name in
            the <tt>Files</tt> field of the <file>.changes</file> file.
            It also gives the default for the same field in the binary
            packages.
          </p>
        </sect1>

        <sect1 id="f-Priority">
          <heading><tt>Priority</tt></heading>

          <p>
            This field represents how important it is that the user
            have the package installed. See <ref id="priorities">.
          </p>

          <p>
            When it appears in the <file>debian/control</file> file,
            it gives the value for the subfield of the same name in
            the <tt>Files</tt> field of the <file>.changes</file> file.
            It also gives the default for the same field in the binary
            packages.
          </p>
        </sect1>

        <sect1 id="f-Package">
          <heading><tt>Package</tt></heading>

          <p>
            The name of the binary package.
          </p>

          <p>
            Binary package names must follow the same syntax and
            restrictions as source package names.  See <ref id="f-Source">
            for the details.
          </p>
        </sect1>

        <sect1 id="f-Architecture">
          <heading><tt>Architecture</tt></heading>

          <p>
            Depending on context and the control file used, the
            <tt>Architecture</tt> field can include the following sets of
            values:
            <list>
                <item>
                  A unique single word identifying a Debian machine
                  architecture as described in <ref id="arch-spec">.
                </item>
                <item>
                  An architecture wildcard identifying a set of Debian
                  machine architectures, see <ref id="arch-wildcard-spec">.
                  <tt>any</tt> matches all Debian machine architectures
                  and is the most frequently used.
                </item>
                <item>
                  <tt>all</tt>, which indicates an
                  architecture-independent package.
                </item>
                <item>
                  <tt>source</tt>, which indicates a source package.
                </item>
            </list>
          </p>

          <p>
            In the main <file>debian/control</file> file in the source
            package, this field may contain the special
            value <tt>all</tt>, the special architecture
            wildcard <tt>any</tt>, or a list of specific and wildcard
            architectures separated by spaces.  If <tt>all</tt>
            or <tt>any</tt> appears, that value must be the entire
            contents of the field.  Most packages will use
            either <tt>all</tt> or <tt>any</tt>.
          </p>

          <p>
            Specifying a specific list of architectures indicates that the
            source will build an architecture-dependent package only on
            architectures included in the list.  Specifying a list of
            architecture wildcards indicates that the source will build an
            architecture-dependent package on only those architectures
            that match any of the specified architecture wildcards.
            Specifying a list of architectures or architecture wildcards
            other than <tt>any</tt> is for the minority of cases where a
            program is not portable or is not useful on some
            architectures.  Where possible, the program should be made
            portable instead.
          </p>

          <p>
            In the Debian source control file <file>.dsc</file>, this
            field contains a list of architectures and architecture
            wildcards separated by spaces. When the list contains the
            architecture wildcard <tt>any</tt>, the only other value
            allowed in the list is <tt>all</tt>.
          </p>

          <p>
            The list may include (or consist solely of) the special
            value <tt>all</tt>.  In other words, in <file>.dsc</file>
            files unlike the <file>debian/control</file>, <tt>all</tt> may
            occur in combination with specific architectures.
            The <tt>Architecture</tt> field in the Debian source control
            file <file>.dsc</file> is generally constructed from
            the <tt>Architecture</tt> fields in
            the <file>debian/control</file> in the source package.
          </p>

          <p>
            Specifying only <tt>any</tt> indicates that the source package
            isn't dependent on any particular architecture and should
            compile fine on any one. The produced binary package(s)
            will be specific to whatever the current build architecture is.
          </p>

          <p>
            Specifying only <tt>all</tt> indicates that the source package
            will only build architecture-independent packages.
          </p>

          <p>
            Specifying <tt>any all</tt> indicates that the source package
            isn't dependent on any particular architecture. The set of
            produced binary packages will include at least one
            architecture-dependant package and one architecture-independent
            package.
          </p>

          <p>
            Specifying a list of architectures or architecture wildcards
            indicates that the source will build an architecture-dependent
            package, and will only work correctly on the listed or
            matching architectures.  If the source package also builds at
            least one architecture-independent package, <tt>all</tt> will
            also be included in the list.
          </p>

          <p>
            In a <file>.changes</file> file, the <tt>Architecture</tt>
            field lists the architecture(s) of the package(s) currently
            being uploaded.  This will be a list; if the source for the
            package is also being uploaded, the special
            entry <tt>source</tt> is also present.  <tt>all</tt> will be
            present if any architecture-independent packages are being
            uploaded.  Architecture wildcards such as <tt>any</tt> must
            never occur in the <tt>Architecture</tt> field in
            the <file>.changes</file> file.
          </p>

          <p>
            See <ref id="debianrules"> for information on how to get
            the architecture for the build process.
          </p>
        </sect1>

        <sect1 id="f-Essential">
          <heading><tt>Essential</tt></heading>

          <p>
            This is a boolean field which may occur only in the
            control file of a binary package or in a per-package fields
            paragraph of a source package control file.
          </p>

          <p>
            If set to <tt>yes</tt> then the package management system
            will refuse to remove the package (upgrading and replacing
            it is still possible). The other possible value is <tt>no</tt>,
            which is the same as not having the field at all.
          </p>
        </sect1>

        <sect1>
          <heading>Package interrelationship fields:
            <tt>Depends</tt>, <tt>Pre-Depends</tt>,
            <tt>Recommends</tt>, <tt>Suggests</tt>,
            <tt>Breaks</tt>, <tt>Conflicts</tt>,
            <tt>Provides</tt>, <tt>Replaces</tt>, <tt>Enhances</tt>
          </heading>

          <p>
            These fields describe the package's relationships with
            other packages.  Their syntax and semantics are described
            in <ref id="relationships">.</p>
        </sect1>

        <sect1 id="f-Standards-Version">
          <heading><tt>Standards-Version</tt></heading>

          <p>
            The most recent version of the standards (the policy
            manual and associated texts) with which the package
            complies.
          </p>

          <p>
            The version number has four components: major and minor
            version number and major and minor patch level.  When the
            standards change in a way that requires every package to
            change the major number will be changed.  Significant
            changes that will require work in many packages will be
            signaled by a change to the minor number.  The major patch
            level will be changed for any change to the meaning of the
            standards, however small; the minor patch level will be
            changed when only cosmetic, typographical or other edits
            are made which neither change the meaning of the document
            nor affect the contents of packages.
          </p>

          <p>
            Thus only the first three components of the policy version
            are significant in the <em>Standards-Version</em> control
            field, and so either these three components or all four
            components may be specified.<footnote>
                In the past, people specified the full version number
                in the Standards-Version field, for example "2.3.0.0".
                Since minor patch-level changes don't introduce new
                policy, it was thought it would be better to relax
                policy and only require the first 3 components to be
                specified, in this example "2.3.0".  All four
                components may still be used if someone wishes to do so.
            </footnote>
          </p>

        </sect1>

        <sect1 id="f-Version">
          <heading><tt>Version</tt></heading>

          <p>
            The version number of a package. The format is:
            [<var>epoch</var><tt>:</tt>]<var>upstream_version</var>[<tt>-</tt><var>debian_revision</var>]
          </p>

          <p>
            The three components here are:
            <taglist>
              <tag><var>epoch</var></tag>
              <item>
                <p>
                  This is a single (generally small) unsigned integer.  It
                  may be omitted, in which case zero is assumed.  If it is
                  omitted then the <var>upstream_version</var> may not
                  contain any colons.
                </p>

                <p>
                  It is provided to allow mistakes in the version numbers
                  of older versions of a package, and also a package's
                  previous version numbering schemes, to be left behind.
                </p>
              </item>

              <tag><var>upstream_version</var></tag>
              <item>
                <p>
                  This is the main part of the version number.  It is
                  usually the version number of the original ("upstream")
                  package from which the <file>.deb</file> file has been made,
                  if this is applicable.  Usually this will be in the same
                  format as that specified by the upstream author(s);
                  however, it may need to be reformatted to fit into the
                  package management system's format and comparison
                  scheme.
                </p>

                <p>
                  The comparison behavior of the package management system
                  with respect to the <var>upstream_version</var> is
                  described below.  The <var>upstream_version</var>
                  portion of the version number is mandatory.
                </p>

                <p>
                  The <var>upstream_version</var> may contain only
                  alphanumerics<footnote>
                        Alphanumerics are <tt>A-Za-z0-9</tt> only.
                  </footnote>
                  and the characters <tt>.</tt> <tt>+</tt> <tt>-</tt>
                  <tt>:</tt> <tt>~</tt> (full stop, plus, hyphen, colon,
                  tilde) and should start with a digit.  If there is no
                  <var>debian_revision</var> then hyphens are not allowed;
                  if there is no <var>epoch</var> then colons are not
                  allowed.
                </p>
              </item>

              <tag><var>debian_revision</var></tag>
              <item>
                <p>
                  This part of the version number specifies the version of
                  the Debian package based on the upstream version.  It
                  may contain only alphanumerics and the characters
                  <tt>+</tt> <tt>.</tt> <tt>~</tt> (plus, full stop,
                  tilde) and is compared in the same way as the
                  <var>upstream_version</var> is.
                </p>

                <p>
                  It is optional; if it isn't present then the
                  <var>upstream_version</var> may not contain a hyphen.
                  This format represents the case where a piece of
                  software was written specifically to be a Debian
                  package, where the Debian package source must always
                  be identical to the pristine source and therefore no
                  revision indication is required.
                </p>

                <p>
                  It is conventional to restart the
                  <var>debian_revision</var> at <tt>1</tt> each time the
                  <var>upstream_version</var> is increased.
                </p>

                <p>
                  The package management system will break the version
                  number apart at the last hyphen in the string (if there
                  is one) to determine the <var>upstream_version</var> and
                  <var>debian_revision</var>.  The absence of a
                  <var>debian_revision</var> is equivalent to a
                  <var>debian_revision</var> of <tt>0</tt>.
                </p>
              </item>
            </taglist>
          </p>

          <p>
            When comparing two version numbers, first the <var>epoch</var>
            of each are compared, then the <var>upstream_version</var> if
            <var>epoch</var> is equal, and then <var>debian_revision</var>
            if <var>upstream_version</var> is also equal.
            <var>epoch</var> is compared numerically.  The
            <var>upstream_version</var> and <var>debian_revision</var>
            parts are compared by the package management system using the
            following algorithm:
          </p>

          <p>
            The strings are compared from left to right.
          </p>

          <p>
            First the initial part of each string consisting entirely of
            non-digit characters is determined.  These two parts (one of
            which may be empty) are compared lexically.  If a difference
            is found it is returned.  The lexical comparison is a
            comparison of ASCII values modified so that all the letters
            sort earlier than all the non-letters and so that a tilde
            sorts before anything, even the end of a part.  For example,
            the following parts are in sorted order from earliest to
            latest: <tt>~~</tt>, <tt>~~a</tt>, <tt>~</tt>, the empty part,
            <tt>a</tt>.<footnote>
              One common use of <tt>~</tt> is for upstream pre-releases.
              For example, <tt>1.0~beta1~svn1245</tt> sorts earlier than
              <tt>1.0~beta1</tt>, which sorts earlier than <tt>1.0</tt>.
            </footnote>
          </p>

          <p>
            Then the initial part of the remainder of each string which
            consists entirely of digit characters is determined.  The
            numerical values of these two parts are compared, and any
            difference found is returned as the result of the comparison.
            For these purposes an empty string (which can only occur at
            the end of one or both version strings being compared) counts
            as zero.
          </p>

          <p>
            These two steps (comparing and removing initial non-digit
            strings and initial digit strings) are repeated until a
            difference is found or both strings are exhausted.
          </p>

          <p>
            Note that the purpose of epochs is to allow us to leave behind
            mistakes in version numbering, and to cope with situations
            where the version numbering scheme changes.  It is
            <em>not</em> intended to cope with version numbers containing
            strings of letters which the package management system cannot
            interpret (such as <tt>ALPHA</tt> or <tt>pre-</tt>), or with
            silly orderings.<footnote>
              The author of this manual has heard of a package whose
              versions went <tt>1.1</tt>, <tt>1.2</tt>, <tt>1.3</tt>,
              <tt>1</tt>, <tt>2.1</tt>, <tt>2.2</tt>, <tt>2</tt> and so
              forth.
            </footnote>
          </p>
        </sect1>

        <sect1 id="f-Description">
          <heading><tt>Description</tt></heading>

          <p>
            In a source or binary control file, the <tt>Description</tt>
            field contains a description of the binary package, consisting
            of two parts, the synopsis or the short description, and the
            long description.  It is a multiline field with the following
            format:
          </p>

          <p>
<example>
        Description: &lt;single line synopsis&gt;
         &lt;extended description over several lines&gt;
</example>
          </p>

          <p>
            The lines in the extended description can have these formats:
          </p>

          <p><list>

            <item>
              Those starting with a single space are part of a paragraph.
              Successive lines of this form will be word-wrapped when
              displayed. The leading space will usually be stripped off.
              The line must contain at least one non-whitespace character.
            </item>

            <item>
              Those starting with two or more spaces. These will be
              displayed verbatim. If the display cannot be panned
              horizontally, the displaying program will line wrap them "hard"
              (i.e., without taking account of word breaks). If it can they
              will be allowed to trail off to the right. None, one or two
              initial spaces may be deleted, but the number of spaces
              deleted from each line will be the same (so that you can have
              indenting work correctly, for example).  The line must
              contain at least one non-whitespace character.
            </item>

            <item>
              Those containing a single space followed by a single full stop
              character. These are rendered as blank lines. This is the
              <em>only</em> way to get a blank line<footnote>
                Completely empty lines will not be rendered as blank lines.
                Instead, they will cause the parser to think you're starting
                a whole new record in the control file, and will therefore
                likely abort with an error.
              </footnote>.
            </item>

            <item>
              Those containing a space, a full stop and some more characters.
              These are for future expansion. Do not use them.
            </item>

          </list></p>

          <p>
            Do not use tab characters.  Their effect is not predictable.
          </p>

          <p>
            See <ref id="descriptions"> for further information on this.
          </p>

          <p>
            In a <file>.changes</file> file, the <tt>Description</tt>
            field contains a summary of the descriptions for the packages
            being uploaded.  For this case, the first line of the field
            value (the part on the same line as <tt>Description:</tt>) is
            always empty.  It is a multiline field, with one
            line per package.  Each line is
            indented by one space and contains the name of a binary
            package, a space, a hyphen (<tt>-</tt>), a space, and the
            short description line from that package.
          </p>
        </sect1>

        <sect1 id="f-Distribution">
          <heading><tt>Distribution</tt></heading>

          <p>
            In a <file>.changes</file> file or parsed changelog output
            this contains the (space-separated) name(s) of the
            distribution(s) where this version of the package should
            be installed.  Valid distributions are determined by the
            archive maintainers.<footnote>
              Example distribution names in the Debian archive used in
              <file>.changes</file> files are:
                <taglist compact="compact">
                  <tag><em>unstable</em></tag>
                  <item>
                    This distribution value refers to the
                    <em>developmental</em> part of the Debian distribution
                    tree.  Most new packages, new upstream versions of
                    packages and bug fixes go into the <em>unstable</em>
                    directory tree.
                  </item>

                  <tag><em>experimental</em></tag>
                  <item>
                    The packages with this distribution value are deemed
                    by their maintainers to be high risk.  Oftentimes they
                    represent early beta or developmental packages from
                    various sources that the maintainers want people to
                    try, but are not ready to be a part of the other parts
                    of the Debian distribution tree.
                  </item>
                </taglist>

                <p>
                  Others are used for updating stable releases or for
                  security uploads.  More information is available in the
                  Debian Developer's Reference, section "The Debian
                  archive".
                </p>
            </footnote>
            The Debian archive software only supports listing a single
            distribution.  Migration of packages to other distributions is
            handled outside of the upload process.
          </p>
        </sect1>

        <sect1 id="f-Date">
          <heading><tt>Date</tt></heading>

          <p>
            This field includes the date the package was built or last
            edited.  It must be in the same format as the <var>date</var>
            in a <file>debian/changelog</file> entry.
          </p>

          <p>
            The value of this field is usually extracted from the
            <file>debian/changelog</file> file - see
            <ref id="dpkgchangelog">).
          </p>
        </sect1>

        <sect1 id="f-Format">
          <heading><tt>Format</tt></heading>

          <p>
            In <qref id="debianchangesfiles"><file>.changes</file></qref>
            files, this field declares the format version of that file.
            The syntax of the field value is the same as that of
            a <qref id="f-Version">package version number</qref> except
            that no epoch or Debian revision is allowed.  The format
            described in this document is <tt>&changesversion;</tt>.
          </p>

          <p>
            In <qref id="debiansourcecontrolfiles"><file>.dsc</file>
            Debian source control</qref> files, this field declares the
            format of the source package.  The field value is used by
            programs acting on a source package to interpret the list of
            files in the source package and determine how to unpack it.
            The syntax of the field value is a numeric major revision, a
            period, a numeric minor revision, and then an optional subtype
            after whitespace, which if specified is an alphanumeric word
            in parentheses.  The subtype is optional in the syntax but may
            be mandatory for particular source format revisions.
            <footnote>
              The source formats currently supported by the Debian archive
              software are <tt>1.0</tt>, <tt>3.0 (native)</tt>,
              and <tt>3.0 (quilt)</tt>.
            </footnote>
          </p>
        </sect1>

        <sect1 id="f-Urgency">
          <heading><tt>Urgency</tt></heading>

          <p>
            This is a description of how important it is to upgrade to
            this version from previous ones.  It consists of a single
            keyword taking one of the values <tt>low</tt>,
            <tt>medium</tt>, <tt>high</tt>, <tt>emergency</tt>, or
            <tt>critical</tt><footnote>
              Other urgency values are supported with configuration
              changes in the archive software but are not used in Debian.
              The urgency affects how quickly a package will be considered
              for inclusion into the <tt>testing</tt> distribution and
              gives an indication of the importance of any fixes included
              in the upload.  <tt>Emergency</tt> and <tt>critical</tt> are
              treated as synonymous.
            </footnote> (not case-sensitive) followed by an optional
            commentary (separated by a space) which is usually in
            parentheses.  For example:

            <example>
  Urgency: low (HIGH for users of diversions)
            </example>

          </p>

          <p>
            The value of this field is usually extracted from the
            <file>debian/changelog</file> file - see
            <ref id="dpkgchangelog">.
          </p>
        </sect1>

        <sect1 id="f-Changes">
          <heading><tt>Changes</tt></heading>

          <p>
            This multiline field contains the human-readable changes data, describing
            the differences between the last version and the current one.
          </p>

          <p>
            The first line of the field value (the part on the same line
            as <tt>Changes:</tt>) is always empty.  The content of the
            field is expressed as continuation lines, with each line
            indented by at least one space.  Blank lines must be
            represented by a line consisting only of a space and a full
            stop (<tt>.</tt>).
          </p>

          <p>
            The value of this field is usually extracted from the
            <file>debian/changelog</file> file - see
            <ref id="dpkgchangelog">).
          </p>

          <p>
            Each version's change information should be preceded by a
            "title" line giving at least the version, distribution(s)
            and urgency, in a human-readable way.
          </p>

          <p>
            If data from several versions is being returned the entry
            for the most recent version should be returned first, and
            entries should be separated by the representation of a
            blank line (the "title" line may also be followed by the
            representation of a blank line).
          </p>
        </sect1>

        <sect1 id="f-Binary">
          <heading><tt>Binary</tt></heading>

          <p>
            This folded field is a list of binary packages.  Its syntax and
            meaning varies depending on the control file in which it
            appears.
          </p>

          <p>
            When it appears in the <file>.dsc</file> file, it lists binary
            packages which a source package can produce, separated by
            commas<footnote>
                A space after each comma is conventional.
            </footnote>.  The source package
            does not necessarily produce all of these binary packages for
            every architecture.  The source control file doesn't contain
            details of which architectures are appropriate for which of
            the binary packages.
          </p>

          <p>
            When it appears in a <file>.changes</file> file, it lists the
            names of the binary packages being uploaded, separated by
            whitespace (not commas).
          </p>
        </sect1>

        <sect1 id="f-Installed-Size">
          <heading><tt>Installed-Size</tt></heading>

          <p>
            This field appears in the control files of binary packages,
            and in the <file>Packages</file> files.  It gives an estimate
            of the total amount of disk space required to install the
            named package.  Actual installed size may vary based on block
            size, file system properties, or actions taken by package
            maintainer scripts.
          </p>

          <p>
            The disk space is given as the integer value of the estimated
            installed size in bytes, divided by 1024 and rounded up.
          </p>
        </sect1>

        <sect1 id="f-Files">
          <heading><tt>Files</tt></heading>

          <p>
            This field contains a list of files with information about
            each one.  The exact information and syntax varies with
            the context.
          </p>

          <p>
            In all cases, Files is a multiline field.  The first line of
            the field value (the part on the same line as <tt>Files:</tt>)
            is always empty.  The content of the field is expressed as
            continuation lines, one line per file.  Each line must be
            indented by one space and contain a number of sub-fields,
            separated by spaces, as described below.
          </p>

          <p>
            In the <file>.dsc</file> file, each line contains the MD5
            checksum, size and filename of the tar file and (if
            applicable) diff file which make up the remainder of the
            source package<footnote>
              That is, the parts which are not the <tt>.dsc</tt>.
            </footnote>.  For example:
            <example>
Files:
 c6f698f19f2a2aa07dbb9bbda90a2754 571925 example_1.2.orig.tar.gz
 938512f08422f3509ff36f125f5873ba 6220 example_1.2-1.diff.gz
            </example>
            The exact forms of the filenames are described
            in <ref id="pkg-sourcearchives">.
          </p>

          <p>
            In the <file>.changes</file> file this contains one line per
            file being uploaded.  Each line contains the MD5 checksum,
            size, section and priority and the filename.  For example:
            <example>
Files:
 4c31ab7bfc40d3cf49d7811987390357 1428 text extra example_1.2-1.dsc
 c6f698f19f2a2aa07dbb9bbda90a2754 571925 text extra example_1.2.orig.tar.gz
 938512f08422f3509ff36f125f5873ba 6220 text extra example_1.2-1.diff.gz
 7c98fe853b3bbb47a00e5cd129b6cb56 703542 text extra example_1.2-1_i386.deb
            </example>
            The <qref id="f-Section">section</qref>
            and <qref id="f-Priority">priority</qref> are the values of
            the corresponding fields in the main source control file.  If
            no section or priority is specified then <tt>-</tt> should be
            used, though section and priority values must be specified for
            new packages to be installed properly.
          </p>

          <p>
            The special value <tt>byhand</tt> for the section in a
            <tt>.changes</tt> file indicates that the file in question
            is not an ordinary package file and must be installed by
            hand by the distribution maintainers.  If the section is
            <tt>byhand</tt> the priority should be <tt>-</tt>.
          </p>

          <p>
            If a new Debian revision of a package is being shipped and
            no new original source archive is being distributed the
            <tt>.dsc</tt> must still contain the <tt>Files</tt> field
            entry for the original source archive
            <file><var>package</var>_<var>upstream-version</var>.orig.tar.gz</file>,
            but the <file>.changes</file> file should leave it out.  In
            this case the original source archive on the distribution
            site must match exactly, byte-for-byte, the original
            source archive which was used to generate the
            <file>.dsc</file> file and diff which are being uploaded.</p>
        </sect1>

        <sect1 id="f-Closes">
          <heading><tt>Closes</tt></heading>

          <p>
            A space-separated list of bug report numbers that the upload
            governed by the .changes file closes.
          </p>
        </sect1>

        <sect1 id="f-Homepage">
          <heading><tt>Homepage</tt></heading>

          <p>
            The URL of the web site for this package, preferably (when
            applicable) the site from which the original source can be
            obtained and any additional upstream documentation or
            information may be found.  The content of this field is a
            simple URL without any surrounding characters such as
            <tt>&lt;&gt;</tt>.
          </p>
        </sect1>

        <sect1 id="f-Checksums">
          <heading><tt>Checksums-Sha1</tt>
            and <tt>Checksums-Sha256</tt></heading>

          <p>
            These multiline fields contain a list of files with a checksum and size
            for each one.  Both <tt>Checksums-Sha1</tt>
            and <tt>Checksums-Sha256</tt> have the same syntax and differ
            only in the checksum algorithm used: SHA-1
            for <tt>Checksums-Sha1</tt> and SHA-256
            for <tt>Checksums-Sha256</tt>.
          </p>

          <p>
            <tt>Checksums-Sha1</tt> and <tt>Checksums-Sha256</tt> are
            multiline fields.  The first line of the field value (the part
            on the same line as <tt>Checksums-Sha1:</tt>
            or <tt>Checksums-Sha256:</tt>) is always empty.  The content
            of the field is expressed as continuation lines, one line per
            file.  Each line consists of the checksum, a space, the file
            size, a space, and the file name.  For example (from
            a <file>.changes</file> file):
            <example>
Checksums-Sha1:
 1f418afaa01464e63cc1ee8a66a05f0848bd155c 1276 example_1.0-1.dsc
 a0ed1456fad61116f868b1855530dbe948e20f06 171602 example_1.0.orig.tar.gz
 5e86ecf0671e113b63388dac81dd8d00e00ef298 6137 example_1.0-1.debian.tar.gz
 71a0ff7da0faaf608481195f9cf30974b142c183 548402 example_1.0-1_i386.deb
Checksums-Sha256:
 ac9d57254f7e835bed299926fd51bf6f534597cc3fcc52db01c4bffedae81272 1276 example_1.0-1.dsc
 0d123be7f51e61c4bf15e5c492b484054be7e90f3081608a5517007bfb1fd128 171602 example_1.0.orig.tar.gz
 f54ae966a5f580571ae7d9ef5e1df0bd42d63e27cb505b27957351a495bc6288 6137 example_1.0-1.debian.tar.gz
 3bec05c03974fdecd11d020fc2e8250de8404867a8a2ce865160c250eb723664 548402 example_1.0-1_i386.deb
            </example>
          </p>

          <p>
            In the <file>.dsc</file> file, these fields list all
            files that make up the source package.  In
            the <file>.changes</file> file, these fields list all
            files being uploaded.  The list of files in these fields
            must match the list of files in the <tt>Files</tt> field.
          </p>
        </sect1>

        <sect1>
          <heading><tt>DM-Upload-Allowed</tt></heading>

          <p>
            Obsolete, see <qref id="f-DM-Upload-Allowed">below</qref>.
          </p>
        </sect1>

        <sect1 id="f-VCS-fields">
          <heading>Version Control System (VCS) fields</heading>

          <p>
            Debian source packages are increasingly developed using VCSs.  The
            purpose of the following fields is to indicate a publicly accessible
            repository where the Debian source package is developed.

            <taglist>
              <tag><tt>Vcs-Browser</tt></tag>
              <item>
                <p>
                  URL of a web interface for browsing the repository.
                </p>
              </item>

              <tag>
                <tt>Vcs-Arch</tt>, <tt>Vcs-Bzr</tt> (Bazaar), <tt>Vcs-Cvs</tt>,
                <tt>Vcs-Darcs</tt>, <tt>Vcs-Git</tt>, <tt>Vcs-Hg</tt>
                (Mercurial), <tt>Vcs-Mtn</tt> (Monotone), <tt>Vcs-Svn</tt>
                (Subversion)
              </tag>
              <item>
                <p>
                  The field name identifies the VCS. The field's value uses the
                  version control system's conventional syntax for describing
                  repository locations and should be sufficient to locate the
                  repository used for packaging. Ideally, it also locates the
                  branch used for development of new versions of the Debian
                  package.
                </p>
                <p>
                  In the case of Git, the value consists of a URL, optionally
                  followed by the word <tt>-b</tt> and the name of a branch in
                  the indicated repository, following the syntax of the
                  <tt>git clone</tt> command.  If no branch is specified, the
                  packaging should be on the default branch.
                </p>
                <p>
                  More than one different VCS may be specified for the same
                  package.
                </p>
              </item>
            </taglist>
          </p>
        </sect1>

        <sect1 id="f-Package-List">
          <heading><tt>Package-List</tt></heading>

          <p>
            Multiline field listing all the packages that can be built from
            the source package, considering every architecture.  The first line
            of the field value is empty.  Each one of the next lines describes
            one binary package, by listing its name, type, section and priority
            separated by spaces.  Fifth and subsequent space-separated items
            may be present and parsers must allow them.  See the
            <qref id="f-Package-Type">Package-Type</qref> field for a list of
            package types.
          </p>
        </sect1>

        <sect1 id="f-Package-Type">
          <heading><tt>Package-Type</tt></heading>

          <p>
            Simple field containing a word indicating the type of package:
            <tt>deb</tt> for binary packages and <tt>udeb</tt> for micro binary
            packages.  Other types not defined here may be indicated.  In
            source package control files, the <tt>Package-Type</tt> field
            should be omitted instead of giving it a value of <tt>deb</tt>, as
            this value is assumed for paragraphs lacking this field.
          </p>
        </sect1>

        <sect1 id="f-Dgit">
          <heading><tt>Dgit</tt></heading>

          <p>
            Folded field containing a single git commit hash, presented in
            full, followed optionally by whitespace and other data to be
            defined in future extensions.
          </p>

          <p>
            Declares that the source package corresponds exactly to a
            referenced commit in a Git repository available at the canonical
            location called <em>dgit-repos</em>, used by <prgn>dgit</prgn>, a
            bidirectional gateway between the Debian archive and Git.  The
            commit is reachable from at least one reference whose name matches
            <tt>refs/dgit/*</tt>.  See the manual page of <prgn>dgit</prgn> for
            further details.
          </p>
        </sect1>
      </sect>

      <sect>
        <heading>User-defined fields</heading>

        <p>
          Additional user-defined fields may be added to the
          source package control file.  Such fields will be
          ignored, and not copied to (for example) binary or
          Debian source control files or upload control files.
        </p>

        <p>
          If you wish to add additional unsupported fields to
          these output files you should use the mechanism
          described here.
        </p>

        <p>
          Fields in the main source control information file with
          names starting <tt>X</tt>, followed by one or more of
          the letters <tt>BCS</tt> and a hyphen <tt>-</tt>, will
          be copied to the output files.  Only the part of the
          field name after the hyphen will be used in the output
          file.  Where the letter <tt>B</tt> is used the field
          will appear in binary package control files, where the
          letter <tt>S</tt> is used in Debian source control
          files and where <tt>C</tt> is used in upload control
          (<tt>.changes</tt>) files.
        </p>

        <p>
          For example, if the main source information control file
          contains the field
          <example>
  XBS-Comment: I stand between the candle and the star.
          </example>
          then the binary and Debian source control files will contain the
          field
          <example>
  Comment: I stand between the candle and the star.
          </example>
        </p>

      </sect>

      <sect id="obsolete-control-data-fields">
        <heading>Obsolete fields</heading>

        <p>
          The following fields have been obsoleted and may be found in packages
          conforming with previous versions of the Policy.
        </p>

        <sect1 id="f-DM-Upload-Allowed">
          <heading><tt>DM-Upload-Allowed</tt></heading>

          <p>
            Indicates that Debian Maintainers may upload this package to
            the Debian archive.  The only valid value is <tt>yes</tt>.  This
            field was used to regulate uploads by Debian Maintainers, See the
            General Resolution <url id="http://www.debian.org/vote/2007/vote_003"
            name="Endorse the concept of Debian Maintainers"> for more details.
          </p>
        </sect1>

      </sect>

    </chapt>


    <chapt id="maintainerscripts">
      <heading>Package maintainer scripts and installation procedure</heading>

      <sect>
        <heading>Introduction to package maintainer scripts</heading>

        <p>
          It is possible to supply scripts as part of a package which
          the package management system will run for you when your
          package is installed, upgraded or removed.
        </p>

        <p>
          These scripts are the control information
          files <prgn>preinst</prgn>, <prgn>postinst</prgn>, <prgn>prerm</prgn>
          and <prgn>postrm</prgn>.  They must be proper executable files;
          if they are scripts (which is recommended), they must start with
          the usual <tt>#!</tt> convention.  They should be readable and
          executable by anyone, and must not be world-writable.
        </p>

        <p>
          The package management system looks at the exit status from
          these scripts.  It is important that they exit with a
          non-zero status if there is an error, so that the package
          management system can stop its processing.  For shell
          scripts this means that you <em>almost always</em> need to
          use <tt>set -e</tt> (this is usually true when writing shell
          scripts, in fact).  It is also important, of course, that
          they exit with a zero status if everything went well.
        </p>

        <p>
          Additionally, packages interacting with users
          using <prgn>debconf</prgn> in the <prgn>postinst</prgn> script
          should install a <prgn>config</prgn> script as a control
          information file.  See <ref id="maintscriptprompt"> for details.
        </p>

        <p>
          When a package is upgraded a combination of the scripts from
          the old and new packages is called during the upgrade
          procedure.  If your scripts are going to be at all
          complicated you need to be aware of this, and may need to
          check the arguments to your scripts.
        </p>

        <p>
          Broadly speaking the <prgn>preinst</prgn> is called before
          (a particular version of) a package is unpacked, and the
          <prgn>postinst</prgn> afterwards; the <prgn>prerm</prgn>
          before (a version of) a package is removed and the
          <prgn>postrm</prgn> afterwards.
        </p>

        <p>
          Programs called from maintainer scripts should not normally
          have a path prepended to them. Before installation is
          started, the package management system checks to see if the
          programs <prgn>ldconfig</prgn>, <prgn>start-stop-daemon</prgn>,
          and <prgn>update-rc.d</prgn> can be found via the
          <tt>PATH</tt> environment variable. Those programs, and any
          other program that one would expect to be in the
          <tt>PATH</tt>, should thus be invoked without an absolute
          pathname. Maintainer scripts should also not reset the
          <tt>PATH</tt>, though they might choose to modify it by
          prepending or appending package-specific directories. These
          considerations really apply to all shell scripts.</p>
      </sect>

      <sect id="idempotency">
        <heading>Maintainer scripts idempotency</heading>

        <p>
          It is necessary for the error recovery procedures that the
          scripts be idempotent.  This means that if it is run
          successfully, and then it is called again, it doesn't bomb
          out or cause any harm, but just ensures that everything is
          the way it ought to be.  If the first call failed, or
          aborted half way through for some reason, the second call
          should merely do the things that were left undone the first
          time, if any, and exit with a success status if everything
          is OK.<footnote>
              This is so that if an error occurs, the user interrupts
              <prgn>dpkg</prgn> or some other unforeseen circumstance
              happens you don't leave the user with a badly-broken
              package when <prgn>dpkg</prgn> attempts to repeat the
              action.
          </footnote>
        </p>
      </sect>

      <sect id="controllingterminal">
        <heading>Controlling terminal for maintainer scripts</heading>

        <p>
          Maintainer scripts are not guaranteed to run with a controlling
          terminal and may not be able to interact with the user.  They
          must be able to fall back to noninteractive behavior if no
          controlling terminal is available.  Maintainer scripts that
          prompt via a program conforming to the Debian Configuration
          Management Specification (see <ref id="maintscriptprompt">) may
          assume that program will handle falling back to noninteractive
          behavior.
        </p>

        <p>
          For high-priority prompts without a reasonable default answer,
          maintainer scripts may abort if there is no controlling
          terminal.  However, this situation should be avoided if at all
          possible, since it prevents automated or unattended installs.
          In most cases, users will consider this to be a bug in the
          package.
        </p>
      </sect>

      <sect id="exitstatus">
        <heading>Exit status</heading>

        <p>
          Each script must return a zero exit status for
          success, or a nonzero one for failure, since the package
          management system looks for the exit status of these scripts
          and determines what action to take next based on that datum.
        </p>
      </sect>

      <sect id="mscriptsinstact"><heading>Summary of ways maintainer
          scripts are called
        </heading>

        <p>
          What follows is a summary of all the ways in which maintainer
          scripts may be called along with what facilities those scripts
          may rely on being available at that time.  Script names preceded
          by <var>new-</var> are the scripts from the new version of a
          package being installed, upgraded to, or downgraded to.  Script
          names preceded by <var>old-</var> are the scripts from the old
          version of a package that is being upgraded from or downgraded
          from.
        </p>

        <p>
          The <prgn>preinst</prgn> script may be called in the following
          ways:
          <taglist>
            <tag><var>new-preinst</var> <tt>install</tt></tag>
            <tag><var>new-preinst</var> <tt>install</tt>
              <var>old-version</var></tag>
            <tag><var>new-preinst</var> <tt>upgrade</tt>
              <var>old-version</var></tag>
            <item>
              The package will not yet be unpacked, so
              the <prgn>preinst</prgn> script cannot rely on any files
              included in its package.  Only essential packages and
              pre-dependencies (<tt>Pre-Depends</tt>) may be assumed to be
              available.  Pre-dependencies will have been configured at
              least once, but at the time the <prgn>preinst</prgn> is
              called they may only be in an "Unpacked" or "Half-Configured"
              state if a previous version of the pre-dependency was
              completely configured and has not been removed since then.
            </item>

            <tag><var>old-preinst</var> <tt>abort-upgrade</tt>
              <var>new-version</var></tag>
            <item>
              Called during error handling of an upgrade that failed after
              unpacking the new package because the <tt>postrm
              upgrade</tt> action failed.  The unpacked files may be
              partly from the new version or partly missing, so the script
              cannot rely on files included in the package.  Package
              dependencies may not be available.  Pre-dependencies will be
              at least "Unpacked" following the same rules as above, except
              they may be only "Half-Installed" if an upgrade of the
              pre-dependency failed.<footnote>
                This can happen if the new version of the package no
                longer pre-depends on a package that had been partially
                upgraded.
              </footnote>
            </item>
          </taglist>
        </p>

        <p>
          The <prgn>postinst</prgn> script may be called in the following
          ways:
          <taglist>
            <tag><var>postinst</var> <tt>configure</tt>
              <var>most-recently-configured-version</var></tag>
            <item>
              The files contained in the package will be unpacked.  All
              package dependencies will at least be "Unpacked".  If there
              are no circular dependencies involved, all package
              dependencies will be configured.  For behavior in the case
              of circular dependencies, see the discussion
              in <ref id="binarydeps">.
            </item>

            <tag><var>old-postinst</var> <tt>abort-upgrade</tt>
              <var>new-version</var></tag>
            <tag><var>conflictor's-postinst</var> <tt>abort-remove</tt>
              <tt>in-favour</tt> <var>package</var>
              <var>new-version</var></tag>
            <tag><var>postinst</var> <tt>abort-remove</tt></tag>
            <tag><var>deconfigured's-postinst</var>
              <tt>abort-deconfigure</tt> <tt>in-favour</tt>
              <var>failed-install-package</var> <var>version</var>
              [<tt>removing</tt> <var>conflicting-package</var>
              <var>version</var>]</tag>
            <item>
              The files contained in the package will be unpacked.  All
              package dependencies will at least be "Half-Installed" and
              will have previously been configured and not removed.
              However, dependencies may not be configured or even fully
              unpacked in some error situations.<footnote>
                For example, suppose packages foo and bar are "Installed"
                with foo depending on bar.  If an upgrade of bar were
                started and then aborted, and then an attempt to remove
                foo failed because its <prgn>prerm</prgn> script failed,
                foo's <tt>postinst abort-remove</tt> would be called with
                bar only "Half-Installed".
              </footnote>
              The <prgn>postinst</prgn> should still attempt any actions
              for which its dependencies are required, since they will
              normally be available, but consider the correct error
              handling approach if those actions fail.  Aborting
              the <prgn>postinst</prgn> action if commands or facilities
              from the package dependencies are not available is often the
              best approach.
            </item>
          </taglist>
        </p>

        <p>
          The <prgn>prerm</prgn> script may be called in the following
          ways:
          <taglist>
            <tag><var>prerm</var> <tt>remove</tt></tag>
            <tag><var>old-prerm</var>
              <tt>upgrade</tt><var>new-version</var></tag>
            <tag><var>conflictor's-prerm</var> <tt>remove</tt>
              <tt>in-favour</tt> <var>package</var>
              <var>new-version</var></tag>
            <tag><var>deconfigured's-prerm</var> <tt>deconfigure</tt>
              <tt>in-favour</tt> <var>package-being-installed</var>
              <var>version</var> [<tt>removing</tt>
              <var>conflicting-package</var> <var>version</var>]</tag>
            <item>
              The package whose <prgn>prerm</prgn> is being called will be
              at least "Half-Installed".  All package dependencies will at
              least be "Half-Installed" and will have previously been
              configured and not removed.  If there was no error, all
              dependencies will at least be "Unpacked", but these actions
              may be called in various error states where dependencies are
              only "Half-Installed" due to a partial upgrade.
            </item>

            <tag><var>new-prerm</var> <tt>failed-upgrade</tt>
              <var>old-version</var></tag>
            <item>
              Called during error handling when <tt>prerm upgrade</tt>
              fails.  The new package will not yet be unpacked, and all
              the same constraints as for <tt>preinst upgrade</tt> apply.
            </item>
          </taglist>
        </p>

        <p>
          The <prgn>postrm</prgn> script may be called in the following
          ways:
          <taglist>
            <tag><var>postrm</var> <tt>remove</tt></tag>
            <tag><var>postrm</var> <tt>purge</tt></tag>
            <tag><var>old-postrm</var> <tt>upgrade</tt>
              <var>new-version</var></tag>
            <tag><var>disappearer's-postrm</var> <tt>disappear</tt>
                <var>overwriter</var> <var>overwriter-version</var></tag>
            <item>
              The <prgn>postrm</prgn> script is called after the package's
              files have been removed or replaced.  The package
              whose <prgn>postrm</prgn> is being called may have
              previously been deconfigured and only be "Unpacked", at which
              point subsequent package changes do not consider its
              dependencies.  Therefore, all <prgn>postrm</prgn> actions
              may only rely on essential packages and must gracefully skip
              any actions that require the package's dependencies if those
              dependencies are unavailable.<footnote>
                This is often done by checking whether the command or
                facility the <prgn>postrm</prgn> intends to call is
                available before calling it.  For example:
<example>
if [ "$1" = purge ] && [ -e /usr/share/debconf/confmodule ]; then
        . /usr/share/debconf/confmodule
        db_purge
fi
</example>
                in <prgn>postrm</prgn> purges the <prgn>debconf</prgn>
                configuration for the package
                if <package>debconf</package> is installed.
              </footnote>
            </item>

            <tag><var>new-postrm</var> <tt>failed-upgrade</tt>
              <var>old-version</var></tag>
            <item>
              Called when the old <tt>postrm upgrade</tt> action fails.
              The new package will be unpacked, but only essential
              packages and pre-dependencies can be relied on.
              Pre-dependencies will either be configured or will be
              "Unpacked" or "Half-Configured" but previously had been
              configured and was never removed.
            </item>

            <tag><var>new-postrm</var> <tt>abort-install</tt></tag>
            <tag><var>new-postrm</var> <tt>abort-install</tt>
              <var>old-version</var></tag>
            <tag><var>new-postrm</var> <tt>abort-upgrade</tt>
              <var>old-version</var></tag>
            <item>
              Called before unpacking the new package as part of the
              error handling of <prgn>preinst</prgn> failures.  May assume
              the same state as <prgn>preinst</prgn> can assume.
            </item>
          </taglist>
        </p>
      </sect>

      <sect id="unpackphase">
        <heading>Details of unpack phase of installation or upgrade</heading>

        <p>
          The procedure on installation/upgrade/overwrite/disappear
          (i.e., when running <tt>dpkg --unpack</tt>, or the unpack
          stage of <tt>dpkg --install</tt>) is as follows.  In each
          case, if a major error occurs (unless listed below) the
          actions are, in general, run backwards - this means that the
          maintainer scripts are run with different arguments in
          reverse order.  These are the "error unwind" calls listed
          below.

          <enumlist>
            <item>
                <enumlist>
                  <item>
                      If a version of the package is already "Installed", call
                      <example compact="compact">
<var>old-prerm</var> upgrade <var>new-version</var>
                      </example>
                  </item>
                  <item>
                      If the script runs but exits with a non-zero
                      exit status, <prgn>dpkg</prgn> will attempt:
                      <example compact="compact">
<var>new-prerm</var> failed-upgrade <var>old-version</var>
                      </example>
                      If this works, the upgrade continues. If this
                      does not work, the error unwind:
                      <example compact="compact">
<var>old-postinst</var> abort-upgrade <var>new-version</var>
                      </example>
                      If this works, then the old-version is
                      "Installed", if not, the old version is in a
                      "Half-Configured" state.
                  </item>
                </enumlist>
            </item>

            <item>
                If a "conflicting" package is being removed at the same time,
                or if any package will be broken (due to <tt>Breaks</tt>):
                <enumlist>
                  <item>
                      If <tt>--auto-deconfigure</tt> is
                      specified, call, for each package to be deconfigured
                      due to <tt>Breaks</tt>:
                      <example compact="compact">
<var>deconfigured's-prerm</var> deconfigure \
  in-favour <var>package-being-installed</var> <var>version</var>
                      </example>
                      Error unwind:
                      <example compact="compact">
<var>deconfigured's-postinst</var> abort-deconfigure \
  in-favour <var>package-being-installed-but-failed</var> <var>version</var>
                      </example>
                      The deconfigured packages are marked as
                      requiring configuration, so that if
                      <tt>--install</tt> is used they will be
                      configured again if possible.
                  </item>
                  <item>
                      If any packages depended on a conflicting
                      package being removed and <tt>--auto-deconfigure</tt> is
                      specified, call, for each such package:
                      <example compact="compact">
<var>deconfigured's-prerm</var> deconfigure \
  in-favour <var>package-being-installed</var> <var>version</var> \
    removing <var>conflicting-package</var> <var>version</var>
                      </example>
                      Error unwind:
                      <example compact="compact">
<var>deconfigured's-postinst</var> abort-deconfigure \
  in-favour <var>package-being-installed-but-failed</var> <var>version</var> \
    removing <var>conflicting-package</var> <var>version</var>
                      </example>
                      The deconfigured packages are marked as
                      requiring configuration, so that if
                      <tt>--install</tt> is used they will be
                      configured again if possible.
                  </item>
                  <item>
                      To prepare for removal of each conflicting package, call:
                      <example compact="compact">
<var>conflictor's-prerm</var> remove \
  in-favour <var>package</var> <var>new-version</var>
                      </example>
                      Error unwind:
                      <example compact="compact">
<var>conflictor's-postinst</var> abort-remove \
  in-favour <var>package</var> <var>new-version</var>
                      </example>
                  </item>
                </enumlist>
            </item>

            <item>
                <enumlist>
                  <item>
                      If the package is being upgraded, call:
                      <example compact="compact">
<var>new-preinst</var> upgrade <var>old-version</var>
                      </example>
                      If this fails, we call:
                      <example>
<var>new-postrm</var> abort-upgrade <var>old-version</var>
                      </example>
                      <enumlist>
                        <item>
                          <p>
                            If that works, then
                            <example>
<var>old-postinst</var> abort-upgrade <var>new-version</var>
                            </example>
                            is called. If this works, then the old version
                            is in an "Installed" state, or else it is left
                            in an "Unpacked" state.
                          </p>
                        </item>
                        <item>
                          <p>
                            If it fails, then the old version is left
                            in an "Half-Installed" state.
                          </p>
                        </item>
                      </enumlist>
                      
                  </item>
                  <item>
                      Otherwise, if the package had some configuration
                      files from a previous version installed (i.e., it
                      is in the "Config-Files" state):
                      <example compact="compact">
<var>new-preinst</var> install <var>old-version</var>
                      </example>
                      Error unwind:
                      <example>
<var>new-postrm</var> abort-install <var>old-version</var>
                      </example>
                      If this fails, the package is left in a
                      "Half-Installed" state, which requires a
                      reinstall. If it works, the packages is left in
                      a "Config-Files" state.
                  </item>
                  <item>
                      Otherwise (i.e., the package was completely purged):
                      <example compact="compact">
<var>new-preinst</var> install
                      </example>
                      Error unwind:
                      <example compact="compact">
<var>new-postrm</var> abort-install
                      </example>
                      If the error-unwind fails, the package is in a
                      "Half-Installed" phase, and requires a
                      reinstall. If the error unwind works, the
                      package is in the "Not-Installed" state.
                  </item>
                </enumlist>
            </item>

            <item>
              <p>
                The new package's files are unpacked, overwriting any
                that may be on the system already, for example any
                from the old version of the same package or from
                another package.  Backups of the old files are kept
                temporarily, and if anything goes wrong the package
                management system will attempt to put them back as
                part of the error unwind.
              </p>

              <p>
                It is an error for a package to contain files which
                are on the system in another package, unless
                <tt>Replaces</tt> is used (see <ref id="replaces">).
                <!--
                The following paragraph is not currently the case:
                Currently the <tt>- - force-overwrite</tt> flag is
                enabled, downgrading it to a warning, but this may not
                always be the case.
                -->
              </p>

              <p>
                It is a more serious error for a package to contain a
                plain file or other kind of non-directory where another
                package has a directory (again, unless
                <tt>Replaces</tt> is used).  This error can be
                overridden if desired using
                <tt>--force-overwrite-dir</tt>, but this is not
                advisable.
              </p>

              <p>
                Packages which overwrite each other's files produce
                behavior which, though deterministic, is hard for the
                system administrator to understand.  It can easily
                lead to "missing" programs if, for example, a package
                is unpacked which overwrites a file from another
                package, and is then removed again.<footnote>
                    Part of the problem is due to what is arguably a
                    bug in <prgn>dpkg</prgn>.
                </footnote>
              </p>

              <p>
                A directory will never be replaced by a symbolic link
                to a directory or vice versa; instead, the existing
                state (symlink or not) will be left alone and
                <prgn>dpkg</prgn> will follow the symlink if there is
                one.
              </p>
            </item>

            <item>
              <p>
                <enumlist>
                  <item>
                      If the package is being upgraded, call
                      <example compact="compact">
<var>old-postrm</var> upgrade <var>new-version</var>
                      </example>
                  </item>
                  <item>
                      If this fails, <prgn>dpkg</prgn> will attempt:
                      <example compact="compact">
<var>new-postrm</var> failed-upgrade <var>old-version</var>
                      </example>
                      If this works, installation continues. If not, 
                      Error unwind:
                      <example compact="compact">
<var>old-preinst</var> abort-upgrade <var>new-version</var>
                      </example>
                      If this fails, the old version is left in a
                      "Half-Installed" state. If it works, dpkg now
                      calls:
                      <example compact="compact">
<var>new-postrm</var> abort-upgrade <var>old-version</var>
                      </example>
                      If this fails, the old version is left in a
                      "Half-Installed" state. If it works, dpkg now
                      calls:
                      <example compact="compact">
<var>old-postinst</var> abort-upgrade <var>new-version</var>
                      </example>
                      If this fails, the old version is in an
                      "Unpacked" state.
                  </item>
                </enumlist>
              </p>

              <p>
                This is the point of no return - if
                <prgn>dpkg</prgn> gets this far, it won't back off
                past this point if an error occurs.  This will
                leave the package in a fairly bad state, which
                will require a successful re-installation to clear
                up, but it's when <prgn>dpkg</prgn> starts doing
                things that are irreversible.
              </p>
            </item>

            <item>
                Any files which were in the old version of the package
                but not in the new are removed.
            </item>

            <item>
                The new file list replaces the old.
            </item>

            <item>
                The new maintainer scripts replace the old.
            </item>

            <item>
                Any packages all of whose files have been overwritten
                during the installation, and which aren't required for
                dependencies, are considered to have been removed.
                For each such package
                <enumlist>
                  <item>
                      <prgn>dpkg</prgn> calls:
                      <example compact="compact">
<var>disappearer's-postrm</var> disappear \
  <var>overwriter</var> <var>overwriter-version</var>
                      </example>
                  </item>
                  <item>
                      The package's maintainer scripts are removed.
                  </item>
                  <item>
                      It is noted in the status database as being in a
                      sane state, namely "Not-Installed" (any conffiles
                      it may have are ignored, rather than being
                      removed by <prgn>dpkg</prgn>).  Note that
                      disappearing packages do not have their prerm
                      called, because <prgn>dpkg</prgn> doesn't know
                      in advance that the package is going to
                      vanish.
                  </item>
                </enumlist>
            </item>

            <item>
                Any files in the package we're unpacking that are also
                listed in the file lists of other packages are removed
                from those lists.  (This will lobotomize the file list
                of the "conflicting" package if there is one.)
            </item>

            <item>
                The backup files made during installation, above, are
                deleted.
            </item>

            <item>
              <p>
                The new package's status is now sane, and recorded as
                "Unpacked".
              </p>

              <p>
                Here is another point of no return - if the
                conflicting package's removal fails we do not unwind
                the rest of the installation; the conflicting package
                is left in a half-removed limbo.
              </p>
            </item>

            <item>
                If there was a conflicting package we go and do the
                removal actions (described below), starting with the
                removal of the conflicting package's files (any that
                are also in the package being unpacked have already
                been removed from the conflicting package's file list,
                and so do not get removed now).
            </item>
          </enumlist>
        </p>
      </sect>

      <sect id="configdetails"><heading>Details of configuration</heading>

        <p>
          When we configure a package (this happens with <tt>dpkg
            --install</tt> and <tt>dpkg --configure</tt>), we first
          update any <tt>conffile</tt>s and then call:
          <example compact="compact">
<var>postinst</var> configure <var>most-recently-configured-version</var>
          </example>
        </p>

        <p>
          No attempt is made to unwind after errors during
          configuration. If the configuration fails, the package is in
          a "Half-Configured" state, and an error message is generated.
        </p>

        <p>
          If there is no most recently configured version
          <prgn>dpkg</prgn> will pass a null argument.
          <footnote>
            <p>
              Historical note: Truly ancient (pre-1997) versions of
              <prgn>dpkg</prgn> passed <tt>&lt;unknown&gt;</tt>
              (including the angle brackets) in this case.  Even older
              ones did not pass a second argument at all, under any
              circumstance.  Note that upgrades using such an old dpkg
              version are unlikely to work for other reasons, even if
              this old argument behavior is handled by your postinst script.
            </p>
          </footnote>     
        </p>
      </sect>

      <sect id="removedetails"><heading>Details of removal and/or
      configuration purging</heading>

        <p>
          <enumlist>
            <item>
              <p>
                <example compact="compact">
<var>prerm</var> remove
                </example>
              </p>
              <p>
                If prerm fails during replacement due to conflict
                <example>
<var>conflictor's-postinst</var> abort-remove \
  in-favour <var>package</var> <var>new-version</var>
                </example>
                Or else we call:
                <example>
<var>postinst</var> abort-remove
                </example>
              </p>
              <p>
                If this fails, the package is in a "Half-Configured"
                state, or else it remains "Installed".
              </p>
            </item>
            <item>
                The package's files are removed (except <tt>conffile</tt>s).
            </item>
            <item>
                <example compact="compact">
<var>postrm</var> remove
                </example>

              <p>
                If it fails, there's no error unwind, and the package is in
                an "Half-Installed" state.
              </p>
            </item>
            <item>
              <p>
                All the maintainer scripts except the <prgn>postrm</prgn>
                are removed.
              </p>

              <p>
                If we aren't purging the package we stop here.  Note
                that packages which have no <prgn>postrm</prgn> and no
                <tt>conffile</tt>s are automatically purged when
                removed, as there is no difference except for the
                <prgn>dpkg</prgn> status.
              </p>
            </item>
            <item>
                The <tt>conffile</tt>s and any backup files
                (<tt>~</tt>-files, <tt>#*#</tt> files,
                <tt>%</tt>-files, <tt>.dpkg-{old,new,tmp}</tt>, etc.)
                are removed.
            </item>
            <item>
              <p>
                <example compact="compact">
<var>postrm</var> purge
                </example>
              </p>
              <p>
                If this fails, the package remains in a "Config-Files"
                state.
              </p>
            </item>
            <item>
                The package's file list is removed.
            </item>
          </enumlist>

        </p>
      </sect>
    </chapt>


    <chapt id="relationships">
      <heading>Declaring relationships between packages</heading>

      <sect id="depsyntax">
        <heading>Syntax of relationship fields</heading>

        <p>
          These fields all have a uniform syntax.  They are a list of
          package names separated by commas.
        </p>

        <p>
          In the <tt>Depends</tt>, <tt>Recommends</tt>,
          <tt>Suggests</tt>, <tt>Pre-Depends</tt>,
          <tt>Build-Depends</tt> and <tt>Build-Depends-Indep</tt>
          control fields of the package, which declare
          dependencies on other packages, the package names listed may
          also include lists of alternative package names, separated
          by vertical bar (pipe) symbols <tt>|</tt>.  In such a case,
          that part of the dependency can be satisfied by any one of
          the alternative packages.
        </p>

        <p>
          All of the fields except for <tt>Provides</tt> may restrict
          their applicability to particular versions of each named
          package.  This is done in parentheses after each individual
          package name; the parentheses should contain a relation from
          the list below followed by a version number, in the format
          described in <ref id="f-Version">.
        </p>

        <p>
          The relations allowed are <tt>&lt;&lt;</tt>, <tt>&lt;=</tt>,
          <tt>=</tt>, <tt>&gt;=</tt> and <tt>&gt;&gt;</tt> for strictly
          earlier, earlier or equal, exactly equal, later or equal and
          strictly later, respectively.  The deprecated
          forms <tt>&lt;</tt> and <tt>&gt;</tt> were confusingly used to
          mean earlier/later or equal, rather than strictly earlier/later,
          and must not appear in new packages (though <prgn>dpkg</prgn>
          still supports them with a warning).
        </p>

        <p>
          Whitespace may appear at any point in the version
          specification subject to the rules in <ref
          id="controlsyntax">, and must appear where it's necessary to
          disambiguate; it is not otherwise significant.  All of the
          relationship fields can only be folded in source package control files.  For
          consistency and in case of future changes to
          <prgn>dpkg</prgn> it is recommended that a single space be
          used after a version relationship and before a version
          number; it is also conventional to put a single space after
          each comma, on either side of each vertical bar, and before
          each open parenthesis.  When opening a continuation line in a relationship field, it
          is conventional to do so after a comma and before the space
          following that comma.
        </p>

        <p>
          For example, a list of dependencies might appear as:
          <example compact="compact">
Package: mutt
Version: 1.3.17-1
Depends: libc6 (>= 2.2.1), exim | mail-transport-agent
          </example>
        </p>

        <p>
          Relationships may be restricted to a certain set of
          architectures.  This is indicated in brackets after each
          individual package name and the optional version specification.
          The brackets enclose a non-empty list of Debian architecture names
          in the format described in <ref id="arch-spec">,
          separated by whitespace.  Exclamation marks may be prepended to
          each of the names.  (It is not permitted for some names to be
          prepended with exclamation marks while others aren't.)
        </p>

        <p>
          For build relationship fields
          (<tt>Build-Depends</tt>, <tt>Build-Depends-Indep</tt>,
          <tt>Build-Conflicts</tt> and <tt>Build-Conflicts-Indep</tt>), if
          the current Debian host architecture is not in this list and
          there are no exclamation marks in the list, or it is in the list
          with a prepended exclamation mark, the package name and the
          associated version specification are ignored completely for the
          purposes of defining the relationships.
        </p>

        <p>
          For example:
          <example compact="compact">
Source: glibc
Build-Depends-Indep: texinfo
Build-Depends: kernel-headers-2.2.10 [!hurd-i386],
  hurd-dev [hurd-i386], gnumach-dev [hurd-i386]
          </example>
          requires <tt>kernel-headers-2.2.10</tt> on all architectures
          other than hurd-i386 and requires <tt>hurd-dev</tt> and
          <tt>gnumach-dev</tt> only on hurd-i386.
        </p>

        <p>
          For binary relationship fields and the <tt>Built-Using</tt>
          field, the architecture restriction
          syntax is only supported in the source package control
          file <file>debian/control</file>.  When the corresponding binary
          package control file is generated, the relationship will either
          be omitted or included without the architecture restriction
          based on the architecture of the binary package.  This means
          that architecture restrictions must not be used in binary
          relationship fields for architecture-independent packages
          (<tt>Architecture: all</tt>).
        </p>

        <p>
          For example:
          <example compact="compact">
Depends: foo [i386], bar [amd64]
          </example>
          becomes <tt>Depends: foo</tt> when the package is built on
          the <tt>i386</tt> architecture, <tt>Depends: bar</tt> when the
          package is built on the <tt>amd64</tt> architecture, and omitted
          entirely in binary packages built on all other architectures.
        </p>

        <p>
          If the architecture-restricted dependency is part of a set of
          alternatives using <tt>|</tt>, that alternative is ignored
          completely on architectures that do not match the restriction.
          For example:
          <example compact="compact">
Build-Depends: foo [!i386] | bar [!amd64]
          </example>
          is equivalent to <tt>bar</tt> on the i386 architecture, to
          <tt>foo</tt> on the amd64 architecture, and to <tt>foo |
          bar</tt> on all other architectures.
        </p>

        <p>
          Relationships may also be restricted to a certain set of
          architectures using architecture wildcards in the format
          described in <ref id="arch-wildcard-spec">.  The syntax for
          declaring such restrictions is the same as declaring
          restrictions using a certain set of architectures without
          architecture wildcards.  For example:
          <example compact="compact">
Build-Depends: foo [linux-any], bar [any-i386], baz [!linux-any]
          </example>
          is equivalent to <tt>foo</tt> on architectures using the Linux
          kernel and any cpu, <tt>bar</tt> on architectures using any
          kernel and an i386 cpu, and <tt>baz</tt> on any architecture
          using a kernel other than Linux.
        </p>

        <p>
          Note that the binary package relationship fields such as
          <tt>Depends</tt> appear in one of the binary package
          sections of the control file, whereas the build-time
          relationships such as <tt>Build-Depends</tt> appear in the
          source package section of the control file (which is the
          first section).
        </p>
      </sect>

      <sect id="binarydeps">
        <heading>Binary Dependencies - <tt>Depends</tt>,
          <tt>Recommends</tt>, <tt>Suggests</tt>, <tt>Enhances</tt>,
          <tt>Pre-Depends</tt>
        </heading>

        <p>
          Packages can declare in their control file that they have
          certain relationships to other packages - for example, that
          they may not be installed at the same time as certain other
          packages, and/or that they depend on the presence of others.
        </p>

        <p>
          This is done using the <tt>Depends</tt>, <tt>Pre-Depends</tt>,
          <tt>Recommends</tt>, <tt>Suggests</tt>, <tt>Enhances</tt>,
          <tt>Breaks</tt> and <tt>Conflicts</tt> control fields.
          <tt>Breaks</tt> is described in <ref id="breaks">, and
          <tt>Conflicts</tt> is described in <ref id="conflicts">.  The
          rest are described below.
        </p>

        <p>
          These seven fields are used to declare a dependency
          relationship by one package on another.  Except for
          <tt>Enhances</tt> and <tt>Breaks</tt>, they appear in the
          depending (binary) package's control file.
          (<tt>Enhances</tt> appears in the recommending package's
          control file, and <tt>Breaks</tt> appears in the version of
          depended-on package which causes the named package to
          break).
        </p>

        <p>
          A <tt>Depends</tt> field takes effect <em>only</em> when a
          package is to be configured.  It does not prevent a package
          being on the system in an unconfigured state while its
          dependencies are unsatisfied, and it is possible to replace
          a package whose dependencies are satisfied and which is
          properly installed with a different version whose
          dependencies are not and cannot be satisfied; when this is
          done the depending package will be left unconfigured (since
          attempts to configure it will give errors) and will not
          function properly.  If it is necessary, a
          <tt>Pre-Depends</tt> field can be used, which has a partial
          effect even when a package is being unpacked, as explained
          in detail below.  (The other three dependency fields,
          <tt>Recommends</tt>, <tt>Suggests</tt> and
          <tt>Enhances</tt>, are only used by the various front-ends
          to <prgn>dpkg</prgn> such as <prgn>apt-get</prgn>,
          <prgn>aptitude</prgn>, and <prgn>dselect</prgn>.)
        </p>

        <p>
          Since <tt>Depends</tt> only places requirements on the order in
          which packages are configured, packages in an installation run
          are usually all unpacked first and all configured later.
          <footnote>
            This approach makes dependency resolution easier.  If two
            packages A and B are being upgraded, the installed package A
            depends on exactly the installed package B, and the new
            package A depends on exactly the new package B (a common
            situation when upgrading shared libraries and their
            corresponding development packages), satisfying the
            dependencies at every stage of the upgrade would be
            impossible.  This relaxed restriction means that both new
            packages can be unpacked together and then configured in their
            dependency order.
          </footnote>
        </p>

        <p>
          If there is a circular dependency among packages being installed
          or removed, installation or removal order honoring the
          dependency order is impossible, requiring the dependency loop be
          broken at some point and the dependency requirements violated
          for at least one package.  Packages involved in circular
          dependencies may not be able to rely on their dependencies being
          configured before they themselves are configured, depending on
          which side of the break of the circular dependency loop they
          happen to be on.  If one of the packages in the loop has
          no <prgn>postinst</prgn> script, then the cycle will be broken
          at that package; this ensures that all <prgn>postinst</prgn>
          scripts are run with their dependencies properly configured if
          this is possible.  Otherwise the breaking point is arbitrary.
          Packages should therefore avoid circular dependencies where
          possible, particularly if they have <prgn>postinst</prgn>
          scripts.
        </p>

        <p>
          The meaning of the five dependency fields is as follows:
          <taglist>
            <tag><tt>Depends</tt></tag>
            <item>
              <p>
                This declares an absolute dependency.  A package will
                not be configured unless all of the packages listed in
                its <tt>Depends</tt> field have been correctly
                configured (unless there is a circular dependency as
                described above).
              </p>

              <p>
                The <tt>Depends</tt> field should be used if the
                depended-on package is required for the depending
                package to provide a significant amount of
                functionality.
              </p>

              <p>
                The <tt>Depends</tt> field should also be used if the
                <prgn>postinst</prgn> or <prgn>prerm</prgn> scripts
                require the depended-on package to be unpacked or
                configured in order to run.  In the case of <tt>postinst
                configure</tt>, the depended-on packages will be unpacked
                and configured first.  (If both packages are involved in a
                dependency loop, this might not work as expected; see the
                explanation a few paragraphs back.)  In the case
                of <prgn>prerm</prgn> or other <prgn>postinst</prgn>
                actions, the package dependencies will normally be at
                least unpacked, but they may be only "Half-Installed" if a
                previous upgrade of the dependency failed.
              </p>

              <p>
                Finally, the <tt>Depends</tt> field should be used if the
                depended-on package is needed by the <prgn>postrm</prgn>
                script to fully clean up after the package removal.  There
                is no guarantee that package dependencies will be
                available when <prgn>postrm</prgn> is run, but the
                depended-on package is more likely to be available if the
                package declares a dependency (particularly in the case
                of <tt>postrm remove</tt>).  The <prgn>postrm</prgn>
                script must gracefully skip actions that require a
                dependency if that dependency isn't available.
              </p>
            </item>

            <tag><tt>Recommends</tt></tag>
            <item>
              <p>
                This declares a strong, but not absolute, dependency.
              </p>

              <p>
                The <tt>Recommends</tt> field should list packages
                that would be found together with this one in all but
                unusual installations.
              </p>
            </item>

            <tag><tt>Suggests</tt></tag>
            <item>
                This is used to declare that one package may be more
                useful with one or more others.  Using this field
                tells the packaging system and the user that the
                listed packages are related to this one and can
                perhaps enhance its usefulness, but that installing
                this one without them is perfectly reasonable.
            </item>

            <tag><tt>Enhances</tt></tag>
            <item>
                This field is similar to Suggests but works in the
                opposite direction. It is used to declare that a
                package can enhance the functionality of another
                package.
            </item>

            <tag><tt>Pre-Depends</tt></tag>
            <item>
              <p>
                This field is like <tt>Depends</tt>, except that it
                also forces <prgn>dpkg</prgn> to complete installation
                of the packages named before even starting the
                installation of the package which declares the
                pre-dependency, as follows:
              </p>

              <p>
                When a package declaring a pre-dependency is about to
                be <em>unpacked</em> the pre-dependency can be
                satisfied if the depended-on package is either fully
                configured, <em>or even if</em> the depended-on
                package(s) are only in the "Unpacked" or the "Half-Configured"
                state, provided that they have been configured
                correctly at some point in the past (and not removed
                or partially removed since).  In this case, both the
                previously-configured and currently "Unpacked" or
                "Half-Configured" versions must satisfy any version
                clause in the <tt>Pre-Depends</tt> field.
              </p>

              <p>
                When the package declaring a pre-dependency is about to
                be <em>configured</em>, the pre-dependency will be treated
                as a normal <tt>Depends</tt>.  It will be considered
                satisfied only if the depended-on package has been
                correctly configured.  However, unlike
                with <tt>Depends</tt>, <tt>Pre-Depends</tt> does not
                permit circular dependencies to be broken.  If a circular
                dependency is encountered while attempting to honor
                <tt>Pre-Depends</tt>, the installation will be aborted.
              </p>

              <p>
                <tt>Pre-Depends</tt> are also required if the
                <prgn>preinst</prgn> script depends on the named package.
                It is best to avoid this situation if possible.
              </p>

              <p>
                <tt>Pre-Depends</tt> should be used sparingly,
                preferably only by packages whose premature upgrade or
                installation would hamper the ability of the system to
                continue with any upgrade that might be in progress.
              </p>

              <p>
                You should not specify a <tt>Pre-Depends</tt> entry for a
                package before this has been discussed on the
                <tt>debian-devel</tt> mailing list and a consensus about
                doing that has been reached.  See <ref id="dependencies">.
              </p>
            </item>
          </taglist>
        </p>

        <p>
          When selecting which level of dependency to use you should
          consider how important the depended-on package is to the
          functionality of the one declaring the dependency.  Some
          packages are composed of components of varying degrees of
          importance.  Such a package should list using
          <tt>Depends</tt> the package(s) which are required by the
          more important components.  The other components'
          requirements may be mentioned as Suggestions or
          Recommendations, as appropriate to the components' relative
          importance.
        </p>
      </sect>

      <sect id="breaks">
        <heading>Packages which break other packages - <tt>Breaks</tt></heading>

        <p>
          When one binary package declares that it breaks another,
          <prgn>dpkg</prgn> will refuse to allow the package which
          declares <tt>Breaks</tt> to be unpacked unless the broken
          package is deconfigured first, and it will refuse to
          allow the broken package to be reconfigured.
        </p>

        <p>
          A package will not be regarded as causing breakage merely
          because its configuration files are still installed; it must
          be at least "Half-Installed".
        </p>

        <p>
          A special exception is made for packages which declare that
          they break their own package name or a virtual package which
          they provide (see below): this does not count as a real
          breakage.
        </p>

        <p>
          Normally a <tt>Breaks</tt> entry will have an "earlier than"
          version clause; such a <tt>Breaks</tt> is introduced in the
          version of an (implicit or explicit) dependency which violates
          an assumption or reveals a bug in earlier versions of the broken
          package, or which takes over a file from earlier versions of the
          package named in <tt>Breaks</tt>.  This use of <tt>Breaks</tt>
          will inform higher-level package management tools that the
          broken package must be upgraded before the new one.
        </p>

        <p>
          If the breaking package also overwrites some files from the
          older package, it should use <tt>Replaces</tt> to ensure this
          goes smoothly.  See <ref id="replaces"> for a full discussion
          of taking over files from other packages, including how to
          use <tt>Breaks</tt> in those cases.
        </p>

        <p>
          Many of the cases where <tt>Breaks</tt> should be used were
          previously handled with <tt>Conflicts</tt>
          because <tt>Breaks</tt> did not yet exist.
          Many <tt>Conflicts</tt> fields should now be <tt>Breaks</tt>.
          See <ref id="conflicts"> for more information about the
          differences.
        </p>
      </sect>

      <sect id="conflicts">
        <heading>Conflicting binary packages - <tt>Conflicts</tt></heading>

        <p>
          When one binary package declares a conflict with another using
          a <tt>Conflicts</tt> field, <prgn>dpkg</prgn> will refuse to
          allow them to be unpacked on the system at the same time.  This
          is a stronger restriction than <tt>Breaks</tt>, which prevents
          the broken package from being configured while the breaking
          package is in the "Unpacked" state but allows both packages to
          be unpacked at the same time.
        </p>

        <p>
          If one package is to be unpacked, the other must be removed
          first.  If the package being unpacked is marked as replacing
          (see <ref id="replaces">, but note that <tt>Breaks</tt> should
          normally be used in this case) the one on the system, or the one
          on the system is marked as deselected, or both packages are
          marked <tt>Essential</tt>, then <prgn>dpkg</prgn> will
          automatically remove the package which is causing the conflict.
          Otherwise, it will halt the installation of the new package with
          an error.  This mechanism is specifically designed to produce an
          error when the installed package is <tt>Essential</tt>, but the
          new package is not.
        </p>

        <p>
          A package will not cause a conflict merely because its
          configuration files are still installed; it must be at least
          "Half-Installed".
        </p>

        <p>
          A special exception is made for packages which declare a
          conflict with their own package name, or with a virtual
          package which they provide (see below): this does not
          prevent their installation, and allows a package to conflict
          with others providing a replacement for it.  You use this
          feature when you want the package in question to be the only
          package providing some feature.
        </p>

        <p>
          Normally, <tt>Breaks</tt> should be used instead
          of <tt>Conflicts</tt> since <tt>Conflicts</tt> imposes a
          stronger restriction on the ordering of package installation or
          upgrade and can make it more difficult for the package manager
          to find a correct solution to an upgrade or installation
          problem.  <tt>Breaks</tt> should be used
          <list>
            <item>when moving a file from one package to another (see
              <ref id="replaces">),</item>
            <item>when splitting a package (a special case of the previous
              one), or</item>
            <item>when the breaking package exposes a bug in or interacts
              badly with particular versions of the broken
              package.</item>
          </list>
          <tt>Conflicts</tt> should be used
          <list>
            <item>when two packages provide the same file and will
              continue to do so,</item>
            <item>in conjunction with <tt>Provides</tt> when only one
              package providing a given virtual facility may be unpacked
              at a time (see <ref id="virtual">),</item>
            <item>in other cases where one must prevent simultaneous
              installation of two packages for reasons that are ongoing
              (not fixed in a later version of one of the packages) or
              that must prevent both packages from being unpacked at the
              same time, not just configured.</item>
          </list>
          Be aware that adding <tt>Conflicts</tt> is normally not the best
          solution when two packages provide the same files.  Depending on
          the reason for that conflict, using alternatives or renaming the
          files is often a better approach.  See, for
          example, <ref id="binaries">.
        </p>

        <p>
          Neither <tt>Breaks</tt> nor <tt>Conflicts</tt> should be used
          unless two packages cannot be installed at the same time or
          installing them both causes one of them to be broken or
          unusable.  Having similar functionality or performing the same
          tasks as another package is not sufficient reason to
          declare <tt>Breaks</tt> or <tt>Conflicts</tt> with that package.
        </p>

        <p>
          A <tt>Conflicts</tt> entry may have an "earlier than" version
          clause if the reason for the conflict is corrected in a later
          version of one of the packages.  However, normally the presence
          of an "earlier than" version clause is a sign
          that <tt>Breaks</tt> should have been used instead.  An "earlier
          than" version clause in <tt>Conflicts</tt>
          prevents <prgn>dpkg</prgn> from upgrading or installing the
          package which declares such a conflict until the upgrade or
          removal of the conflicted-with package has been completed, which
          is a strong restriction.
        </p>
      </sect>

      <sect id="virtual"><heading>Virtual packages - <tt>Provides</tt>
        </heading>

        <p>
          As well as the names of actual ("concrete") packages, the
          package relationship fields <tt>Depends</tt>,
          <tt>Recommends</tt>, <tt>Suggests</tt>, <tt>Enhances</tt>,
          <tt>Pre-Depends</tt>, <tt>Breaks</tt>, <tt>Conflicts</tt>,
          <tt>Build-Depends</tt>, <tt>Build-Depends-Indep</tt>,
          <tt>Build-Conflicts</tt> and <tt>Build-Conflicts-Indep</tt>
          may mention "virtual packages".
        </p>

        <p>
          A <em>virtual package</em> is one which appears in the
          <tt>Provides</tt> control field of another package.  The effect
          is as if the package(s) which provide a particular virtual
          package name had been listed by name everywhere the virtual
          package name appears. (See also <ref id="virtual_pkg">)
        </p>

        <p>
          If there are both concrete and virtual packages of the same
          name, then the dependency may be satisfied (or the conflict
          caused) by either the concrete package with the name in
          question or any other concrete package which provides the
          virtual package with the name in question.  This is so that,
          for example, supposing we have
          <example compact="compact">
Package: foo
Depends: bar
          </example> and someone else releases an enhanced version of
          the <tt>bar</tt> package they can say:
          <example compact="compact">
Package: bar-plus
Provides: bar
          </example>
          and the <tt>bar-plus</tt> package will now also satisfy the
          dependency for the <tt>foo</tt> package.
        </p>

        <p>
          If a relationship field has a version number attached, only real
          packages will be considered to see whether the relationship is
          satisfied (or the prohibition violated, for a conflict or
          breakage).  In other words, if a version number is specified,
          this is a request to ignore all <tt>Provides</tt> for that
          package name and consider only real packages.  The package
          manager will assume that a package providing that virtual
          package is not of the "right" version.  A <tt>Provides</tt>
          field may not contain version numbers, and the version number of
          the concrete package which provides a particular virtual package
          will not be considered when considering a dependency on or
          conflict with the virtual package name.<footnote>
            It is possible that a future release of <prgn>dpkg</prgn> may
            add the ability to specify a version number for each virtual
            package it provides.  This feature is not yet present,
            however, and is expected to be used only infrequently.
          </footnote>
        </p>

        <p>
          To specify which of a set of real packages should be the default
          to satisfy a particular dependency on a virtual package, list
          the real package as an alternative before the virtual one.
        </p>

        <p>
          If the virtual package represents a facility that can only be
          provided by one real package at a time, such as
          the <package>mail-transport-agent</package> virtual package that
          requires installation of a binary that would conflict with all
          other providers of that virtual package (see
          <ref id="mail-transport-agents">), all packages providing that
          virtual package should also declare a conflict with it
          using <tt>Conflicts</tt>.  This will ensure that at most one
          provider of that virtual package is unpacked or installed at a
          time.
        </p>
      </sect>

      <sect id="replaces"><heading>Overwriting files and replacing
          packages - <tt>Replaces</tt></heading>

        <p>
          Packages can declare in their control file that they should
          overwrite files in certain other packages, or completely replace
          other packages. The <tt>Replaces</tt> control field has these
          two distinct purposes.
        </p>

        <sect1><heading>Overwriting files in other packages</heading>

          <p>
            It is usually an error for a package to contain files which
            are on the system in another package.  However, if the
            overwriting package declares that it <tt>Replaces</tt> the one
            containing the file being overwritten, then <prgn>dpkg</prgn>
            will replace the file from the old package with that from the
            new.  The file will no longer be listed as "owned" by the old
            package and will be taken over by the new package.
            Normally, <tt>Breaks</tt> should be used in conjunction
            with <tt>Replaces</tt>.<footnote>
              To see why <tt>Breaks</tt> is normally needed in addition
              to <tt>Replaces</tt>, consider the case of a file in the
              package <package>foo</package> being taken over by the
              package <package>foo-data</package>.
              <tt>Replaces</tt> will allow <package>foo-data</package> to
              be installed and take over that file.  However,
              without <tt>Breaks</tt>, nothing
              requires <package>foo</package> to be upgraded to a newer
              version that knows it does not include that file and instead
              depends on <package>foo-data</package>.  Nothing would
              prevent the new <package>foo-data</package> package from
              being installed and then removed, removing the file that it
              took over from <package>foo</package>.  After that
              operation, the package manager would think the system was in
              a consistent state, but the <package>foo</package> package
              would be missing one of its files.
            </footnote>
          </p>

          <p>
            For example, if a package <package>foo</package> is split
            into <package>foo</package> and <package>foo-data</package>
            starting at version 1.2-3, <package>foo-data</package> would
            have the fields
            <example compact="compact">
Replaces: foo (&lt;&lt; 1.2-3)
Breaks: foo (&lt;&lt; 1.2-3)
            </example>
            in its control file.  The new version of the
            package <package>foo</package> would normally have the field
            <example compact="compact">
Depends: foo-data (&gt;= 1.2-3)
            </example>
            (or possibly <tt>Recommends</tt> or even <tt>Suggests</tt> if
            the files moved into <package>foo-data</package> are not
            required for normal operation).
          </p>

          <p>
            If a package is completely replaced in this way, so that
            <prgn>dpkg</prgn> does not know of any files it still
            contains, it is considered to have "disappeared".  It will
            be marked as not wanted on the system (selected for
            removal) and "Not-Installed".  Any <tt>conffile</tt>s
            details noted for the package will be ignored, as they
            will have been taken over by the overwriting package.  The
            package's <prgn>postrm</prgn> script will be run with a
            special argument to allow the package to do any final
            cleanup required.  See <ref id="mscriptsinstact">.
            <footnote>
              Replaces is a one way relationship.  You have to install
              the replacing package after the replaced package.
            </footnote>
          </p>

          <p>
            For this usage of <tt>Replaces</tt>, virtual packages (see
            <ref id="virtual">) are not considered when looking at a
            <tt>Replaces</tt> field.  The packages declared as being
            replaced must be mentioned by their real names.
          </p>

          <p>
            This usage of <tt>Replaces</tt> only takes effect when both
            packages are at least partially on the system at once.  It is
            not relevant if the packages conflict unless the conflict has
            been overridden.
          </p>
        </sect1>

        <sect1><heading>Replacing whole packages, forcing their
            removal</heading>

          <p>
            Second, <tt>Replaces</tt> allows the packaging system to
            resolve which package should be removed when there is a
            conflict (see <ref id="conflicts">).  This usage only takes
            effect when the two packages <em>do</em> conflict, so that the
            two usages of this field do not interfere with each other.
          </p>

          <p>
            In this situation, the package declared as being replaced
            can be a virtual package, so for example, all mail
            transport agents (MTAs) would have the following fields in
            their control files:
            <example compact="compact">
Provides: mail-transport-agent
Conflicts: mail-transport-agent
Replaces: mail-transport-agent
            </example>
            ensuring that only one MTA can be unpacked at any one
            time.  See <ref id="virtual"> for more information about this
            example.
        </sect1>
      </sect>

      <sect id="sourcebinarydeps">
        <heading>Relationships between source and binary packages -
          <tt>Build-Depends</tt>, <tt>Build-Depends-Indep</tt>,
          <tt>Build-Conflicts</tt>, <tt>Build-Conflicts-Indep</tt>
        </heading>

        <p>
          Source packages that require certain binary packages to be
          installed or absent at the time of building the package
          can declare relationships to those binary packages.
        </p>

        <p>
          This is done using the <tt>Build-Depends</tt>,
          <tt>Build-Depends-Indep</tt>, <tt>Build-Conflicts</tt> and
          <tt>Build-Conflicts-Indep</tt> control fields.
        </p>

        <p>
          Build-dependencies on "build-essential" binary packages can be
          omitted. Please see <ref id="pkg-relations"> for more information.
        </p>

        <p>
          The dependencies and conflicts they define must be satisfied
          (as defined earlier for binary packages) in order to invoke
          the targets in <tt>debian/rules</tt>, as follows:<footnote>
            <p>
              There is no Build-Depends-Arch; this role is essentially
              met with Build-Depends.  Anyone building the
              <tt>build-indep</tt> and <tt>binary-indep</tt> targets is
              assumed to be building the whole package, and therefore
              installation of all build dependencies is required.
            </p>
            <p>
              The autobuilders use <tt>dpkg-buildpackage -B</tt>, which
              calls <tt>build</tt>, not <tt>build-arch</tt> since it does
              not yet know how to check for its existence, and
              <tt>binary-arch</tt>.  The purpose of the original split
              between <tt>Build-Depends</tt> and
              <tt>Build-Depends-Indep</tt> was so that the autobuilders
              wouldn't need to install extra packages needed only for the
              binary-indep targets.  But without a build-arch/build-indep
              split, this didn't work, since most of the work is done in
              the build target, not in the binary target.
            </p>
          </footnote>
          <taglist>
            <tag><tt>clean</tt>, <tt>build-arch</tt>, and
              <tt>binary-arch</tt></tag>
            <item>
              Only the <tt>Build-Depends</tt> and <tt>Build-Conflicts</tt>
              fields must be satisfied when these targets are invoked.
            </item>
            <tag><tt>build</tt>, <tt>build-indep</tt>, <tt>binary</tt>,
              and <tt>binary-indep</tt></tag>
            <item>
              The <tt>Build-Depends</tt>, <tt>Build-Conflicts</tt>,
              <tt>Build-Depends-Indep</tt>, and
              <tt>Build-Conflicts-Indep</tt> fields must be satisfied when
              these targets are invoked.
            </item>
          </taglist>
        </p>
      </sect>

      <sect id="built-using">
        <heading>Additional source packages used to build the binary
          - <tt>Built-Using</tt>
        </heading>

        <p>
          Some binary packages incorporate parts of other packages when built
          but do not have to depend on those packages.  Examples include
          linking with static libraries or incorporating source code from
          another package during the build.  In this case, the source packages
          of those other packages are a required part of the complete source
          (the binary package is not reproducible without them).
        </p>

        <p>
          A <tt>Built-Using</tt> field must list the corresponding source
          package for any such binary package incorporated during the build
          <footnote>
            <tt>Build-Depends</tt> in the source package is not adequate since
            it (rightfully) does not document the exact version used in the
            build.
          </footnote>,
          including an "exactly equal" ("=") version relation on the version
          that was used to build that binary package<footnote>
            The archive software might reject packages that refer to
            non-existent sources.
          </footnote>.
        </p>

        <p>
          A package using the source code from the gcc-4.6-source
          binary package built from the gcc-4.6 source package would
          have this field in its control file:
          <example compact="compact">
Built-Using: gcc-4.6 (= 4.6.0-11)
          </example>
        </p>

        <p>
          A package including binaries from grub2 and loadlin would
          have this field in its control file:
          <example compact="compact">
Built-Using: grub2 (= 1.99-9), loadlin (= 1.6e-1)
          </example>
        </p>
      </sect>
    </chapt>


    <chapt id="sharedlibs"><heading>Shared libraries</heading>

      <p>
        Packages containing shared libraries must be constructed with
        a little care to make sure that the shared library is always
        available.  This is especially important for packages whose
        shared libraries are vitally important, such as the C library
        (currently <tt>libc6</tt>).
      </p>

      <p>
        This section deals only with public shared libraries: shared
        libraries that are placed in directories searched by the dynamic
        linker by default or which are intended to be linked against
        normally and possibly used by other, independent packages.  Shared
        libraries that are internal to a particular package or that are
        only loaded as dynamic modules are not covered by this section and
        are not subject to its requirements.
      </p>

      <p>
        A shared library is identified by the <tt>SONAME</tt> attribute
        stored in its dynamic section.  When a binary is linked against a
        shared library, the <tt>SONAME</tt> of the shared library is
        recorded in the binary's <tt>NEEDED</tt> section so that the
        dynamic linker knows that library must be loaded at runtime.  The
        shared library file's full name (which usually contains additional
        version information not needed in the <tt>SONAME</tt>) is
        therefore normally not referenced directly.  Instead, the shared
        library is loaded by its <tt>SONAME</tt>, which exists on the file
        system as a symlink pointing to the full name of the shared
        library.  This symlink must be provided by the
        package.  <ref id="sharedlibs-runtime"> describes how to do this.
        <footnote>
          This is a convention of shared library versioning, but not a
          requirement.  Some libraries use the <tt>SONAME</tt> as the full
          library file name instead and therefore do not need a symlink.
          Most, however, encode additional information about
          backwards-compatible revisions as a minor version number in the
          file name.  The <tt>SONAME</tt> itself only changes when
          binaries linked with the earlier version of the shared library
          may no longer work, but the filename may change with each
          release of the library.  See <ref id="sharedlibs-runtime"> for
          more information.
        </footnote>
      </p>

      <p>
        When linking a binary or another shared library against a shared
        library, the <tt>SONAME</tt> for that shared library is not yet
        known.  Instead, the shared library is found by looking for a file
        matching the library name with <tt>.so</tt> appended.  This file
        exists on the file system as a symlink pointing to the shared
        library.
      </p>

      <p>
        Shared libraries are normally split into several binary packages.
        The <tt>SONAME</tt> symlink is installed by the runtime shared
        library package, and the bare <tt>.so</tt> symlink is installed in
        the development package since it's only used when linking binaries
        or shared libraries.  However, there are some exceptions for
        unusual shared libraries or for shared libraries that are also
        loaded as dynamic modules by other programs.
      </p>

      <p>
        This section is primarily concerned with how the separation of
        shared libraries into multiple packages should be done and how
        dependencies on and between shared library binary packages are
        managed in Debian.  <ref id="libraries"> should be read in
        conjunction with this section and contains additional rules for
        the files contained in the shared library packages.
      </p>

      <sect id="sharedlibs-runtime">
        <heading>Run-time shared libraries</heading>

        <p>
          The run-time shared library must be placed in a package
          whose name changes whenever the <tt>SONAME</tt> of the shared
          library changes.  This allows several versions of the shared
          library to be installed at the same time, allowing installation
          of the new version of the shared library without immediately
          breaking binaries that depend on the old version.  Normally, the
          run-time shared library and its <tt>SONAME</tt> symlink should
          be placed in a package named
          <package><var>libraryname</var><var>soversion</var></package>,
          where <var>soversion</var> is the version number in
          the <tt>SONAME</tt> of the shared library.  Alternatively, if it
          would be confusing to directly append <var>soversion</var>
          to <var>libraryname</var> (if, for
          example, <var>libraryname</var> itself ends in a number), you
          should use
          <package><var>libraryname</var>-<var>soversion</var></package>
          instead.
        </p>

        <p>
          To determine the <var>soversion</var>, look at
          the <tt>SONAME</tt> of the library, stored in the
          ELF <tt>SONAME</tt> attribute.  It is usually of the
          form <tt><var>name</var>.so.<var>major-version</var></tt> (for
          example, <tt>libz.so.1</tt>).  The version part is the part
          which comes after <tt>.so.</tt>, so in that example it
          is <tt>1</tt>.  The soname may instead be of the
          form <tt><var>name</var>-<var>major-version</var>.so</tt>, such
          as <tt>libdb-5.1.so</tt>, in which case the name would
          be <tt>libdb</tt> and the version would be <tt>5.1</tt>.
        </p>

        <p>
          If you have several shared libraries built from the same source
          tree, you may lump them all together into a single shared
          library package provided that all of their <tt>SONAME</tt>s will
          always change together.  Be aware that this is not normally the
          case, and if the <tt>SONAME</tt>s do not change together,
          upgrading such a merged shared library package will be
          unnecessarily difficult because of file conflicts with the old
          version of the package.  When in doubt, always split shared
          library packages so that each binary package installs a single
          shared library.
        </p>

        <p>
          Every time the shared library ABI changes in a way that may
          break binaries linked against older versions of the shared
          library, the <tt>SONAME</tt> of the library and the
          corresponding name for the binary package containing the runtime
          shared library should change.  Normally, this means
          the <tt>SONAME</tt> should change any time an interface is
          removed from the shared library or the signature of an interface
          (the number of parameters or the types of parameters that it
          takes, for example) is changed.  This practice is vital to
          allowing clean upgrades from older versions of the package and
          clean transitions between the old ABI and new ABI without having
          to upgrade every affected package simultaneously.
        </p>

        <p>
          The <tt>SONAME</tt> and binary package name need not, and indeed
          normally should not, change if new interfaces are added but none
          are removed or changed, since this will not break binaries
          linked against the old shared library.  Correct versioning of
          dependencies on the newer shared library by binaries that use
          the new interfaces is handled via
          the <qref id="sharedlibs-depends"><tt>symbols</tt>
          or <tt>shlibs</tt> system</qref>.
        </p>

      <p>
        The package should install the shared libraries under
        their normal names.  For example, the <package>libgdbm3</package>
        package should install <file>libgdbm.so.3.0.0</file> as
        <file>/usr/lib/libgdbm.so.3.0.0</file>.  The files should not be
        renamed or re-linked by any <prgn>prerm</prgn> or
        <prgn>postrm</prgn> scripts; <prgn>dpkg</prgn> will take care
        of renaming things safely without affecting running programs,
        and attempts to interfere with this are likely to lead to
        problems.
      </p>

      <p>
        Shared libraries should not be installed executable, since
        the dynamic linker does not require this and trying to
        execute a shared library usually results in a core dump.
      </p>

      <p>
        The run-time library package should include the symbolic link for
        the <tt>SONAME</tt> that <prgn>ldconfig</prgn> would create for
        the shared libraries.  For example,
        the <package>libgdbm3</package> package should include a symbolic
        link from <file>/usr/lib/libgdbm.so.3</file> to
        <file>libgdbm.so.3.0.0</file>.  This is needed so that the dynamic
        linker (for example <prgn>ld.so</prgn> or
        <prgn>ld-linux.so.*</prgn>) can find the library between the
        time that <prgn>dpkg</prgn> installs it and the time that
        <prgn>ldconfig</prgn> is run in the <prgn>postinst</prgn>
        script.<footnote>
            The package management system requires the library to be
            placed before the symbolic link pointing to it in the
            <file>.deb</file> file.  This is so that when
            <prgn>dpkg</prgn> comes to install the symlink
            (overwriting the previous symlink pointing at an older
            version of the library), the new shared library is already
            in place.  In the past, this was achieved by creating the
            library in the temporary packaging directory before
            creating the symlink.  Unfortunately, this was not always
            effective, since the building of the tar file in the
            <file>.deb</file> depended on the behavior of the underlying
            file system.  Some file systems (such as reiserfs) reorder
            the files so that the order of creation is forgotten.
            Since version 1.7.0, <prgn>dpkg</prgn>
            reorders the files itself as necessary when building a
            package.  Thus it is no longer important to concern
            oneself with the order of file creation.
        </footnote>
      </p>

        <sect1 id="ldconfig">
          <heading><tt>ldconfig</tt></heading>

        <p>
          Any package installing shared libraries in one of the default
          library directories of the dynamic linker (which are currently
          <file>/usr/lib</file> and <file>/lib</file>) or a directory that is
          listed in <file>/etc/ld.so.conf</file><footnote>
            These are currently <file>/usr/local/lib</file> plus
            directories under <file>/lib</file> and <file>/usr/lib</file>
            matching the multiarch triplet for the system architecture.
          </footnote>
          must use <prgn>ldconfig</prgn> to update the shared library
          system.
        </p>

        <p>
            The package maintainer scripts must only call
            <prgn>ldconfig</prgn> under these circumstances:
            <list compact="compact">
              <item>When the <prgn>postinst</prgn> script is run with a
                  first argument of <tt>configure</tt>, the script must call
                  <prgn>ldconfig</prgn>, and may optionally invoke
                  <prgn>ldconfig</prgn> at other times.
              </item>
              <item>When the <prgn>postrm</prgn> script is run with a
                  first argument of <tt>remove</tt>, the script should call
                  <prgn>ldconfig</prgn>.
              </item>
            </list>
         <footnote>
            <p>
              During install or upgrade, the preinst is called before
              the new files are unpacked, so calling "ldconfig" is
              pointless.  The preinst of an existing package can also be
              called if an upgrade fails.  However, this happens during
              the critical time when a shared libs may exist on-disk
              under a temporary name.  Thus, it is dangerous and
              forbidden by current policy to call "ldconfig" at this
              time.
            </p>

            <p>
              When a package is installed or upgraded, "postinst
              configure" runs after the new files are safely on-disk.
              Since it is perfectly safe to invoke ldconfig
              unconditionally in a postinst, it is OK for a package to
              simply put ldconfig in its postinst without checking the
              argument.  The postinst can also be called to recover from
              a failed upgrade.  This happens before any new files are
              unpacked, so there is no reason to call "ldconfig" at this
              point.
            </p>

            <p>
              For a package that is being removed, prerm is
              called with all the files intact, so calling ldconfig is
              useless.  The other calls to "prerm" happen in the case of
              upgrade at a time when all the files of the old package
              are on-disk, so again calling "ldconfig" is pointless.
            </p>

            <p>
              postrm, on the other hand, is called with the "remove"
              argument just after the files are removed, so this is
              the proper time to call "ldconfig" to notify the system
              of the fact that the shared libraries from the package
              are removed.  The postrm can be called at several other
              times.  At the time of "postrm purge", "postrm
              abort-install", or "postrm abort-upgrade", calling
              "ldconfig" is useless because the shared lib files are
              not on-disk.  However, when "postrm" is invoked with
              arguments "upgrade", "failed-upgrade", or "disappear", a
              shared lib may exist on-disk under a temporary filename.
            </p>
          </footnote>
        </p>
        </sect1>

      </sect>

      <sect id="sharedlibs-support-files">
        <heading>Shared library support files</heading>

        <p>
          If your package contains files whose names do not change with
          each change in the library shared object version, you must not
          put them in the shared library package.  Otherwise, several
          versions of the shared library cannot be installed at the same
          time without filename clashes, making upgrades and transitions
          unnecessarily difficult.
        </p>

        <p>
          It is recommended that supporting files and run-time support
          programs that do not need to be invoked manually by users, but
          are nevertheless required for the package to function, be placed
          (if they are binary) in a subdirectory of <file>/usr/lib</file>,
          preferably under <file>/usr/lib/</file><var>package-name</var>.
          If the program or file is architecture independent, the
          recommendation is for it to be placed in a subdirectory of
          <file>/usr/share</file> instead, preferably under
          <file>/usr/share/</file><var>package-name</var>.  Following the
          <var>package-name</var> naming convention ensures that the file
          names change when the shared object version changes.
        </p>

        <p>
          Run-time support programs that use the shared library but are
          not required for the library to function or files used by the
          shared library that can be used by any version of the shared
          library package should instead be put in a separate package.
          This package might typically be named
          <package><var>libraryname</var>-tools</package>; note the
          absence of the <var>soversion</var> in the package name.
        </p>

        <p>
          Files and support programs only useful when compiling software
          against the library should be included in the development
          package for the library.<footnote>
            For example, a <file><var>package-name</var>-config</file>
            script or <package>pkg-config</package> configuration files.
          </footnote>
        </p>
      </sect>

      <sect id="sharedlibs-static">
        <heading>Static libraries</heading>

      <p>
        The static library (<file><var>libraryname.a</var></file>)
        is usually provided in addition to the shared version.
        It is placed into the development package (see below).
      </p>

      <p>
        In some cases, it is acceptable for a library to be
        available in static form only; these cases include:
        <list>
          <item>libraries for languages whose shared library support
                is immature or unstable</item>
          <item>libraries whose interfaces are in flux or under
                development (commonly the case when the library's
                major version number is zero, or where the ABI breaks
                across patchlevels)</item>
          <item>libraries which are explicitly intended to be
                available only in static form by their upstream
                author(s)</item>
        </list>
      </p>

      <sect id="sharedlibs-dev">
        <heading>Development files</heading>

      <p>
        If there are development files associated with a shared library,
        the source package needs to generate a binary development package
        named <package><var>libraryname</var><var>soversion</var>-dev</package>,
        or if you prefer only to support one development version at a
        time, <package><var>libraryname</var>-dev</package>.  Installing
        the development package must result in installation of all the
        development files necessary for compiling programs against that
        shared library.<footnote>
          This wording allows the development files to be split into
          several packages, such as a separate architecture-independent
          <package><var>libraryname</var>-headers</package>, provided that
          the development package depends on all the required additional
          packages.
        </footnote>
      </p>

      <p>
        In case several development versions of a library exist, you may
        need to use <prgn>dpkg</prgn>'s Conflicts mechanism (see
        <ref id="conflicts">) to ensure that the user only installs one
        development version at a time (as different development versions are
        likely to have the same header files in them, which would cause a
        filename clash if both were unpacked).
      </p>

      <p>
        The development package should contain a symlink for the associated
        shared library without a version number. For example, the
        <package>libgdbm-dev</package> package should include a symlink
        from <file>/usr/lib/libgdbm.so</file> to
        <file>libgdbm.so.3.0.0</file>.  This symlink is needed by the linker
        (<prgn>ld</prgn>) when compiling packages, as it will only look for
        <file>libgdbm.so</file> when compiling dynamically.
      </p>

      <p>
        If the package provides Ada Library Information
        (<file>*.ali</file>) files for use with GNAT, these files must be
        installed read-only (mode 0444) so that GNAT will not attempt to
        recompile them.  This overrides the normal file mode requirements
        given in <ref id="permissions-owners">.
      </p>
      </sect>

      <sect id="sharedlibs-intradeps">
        <heading>Dependencies between the packages of the same library</heading>

        <p>
          Typically the development version should have an exact
          version dependency on the runtime library, to make sure that
          compilation and linking happens correctly.  The
          <tt>${binary:Version}</tt> substitution variable can be
          useful for this purpose.
          <footnote>
            Previously, <tt>${Source-Version}</tt> was used, but its name
            was confusing and it has been deprecated since dpkg 1.13.19.
          </footnote>
        </p>
      </sect>

      <sect id="sharedlibs-depends">
        <heading>Dependencies between the library and other
          packages</heading>

        <p>
          If a package contains a binary or library which links to a
          shared library, we must ensure that, when the package is
          installed on the system, all of the libraries needed are also
          installed.  These dependencies must be added to the binary
          package when it is built, since they may change based on which
          version of a shared library the binary or library was linked
          with even if there are no changes to the source of the binary
          (for example, symbol versions change, macros become functions or
          vice versa, or the binary package may determine at compile-time
          whether new library interfaces are available and can be called).
          To allow these dependencies to be constructed, shared libraries
          must provide either a <file>symbols</file> file or
          a <file>shlibs</file> file.  These provide information on the
          package dependencies required to ensure the presence of
          interfaces provided by this library.  Any package with binaries
          or libraries linking to a shared library must use these files to
          determine the required dependencies when it is built.  Other
          packages which use a shared library (for example using
          <tt>dlopen()</tt>) should compute appropriate dependencies
          using these files at build time as well.
        </p>

        <p>
          The two mechanisms differ in the degree of detail that they
          provide.  A <file>symbols</file> file documents, for each symbol
          exported by a library, the minimal version of the package any
          binary using this symbol will need.  This is typically the
          version of the package in which the symbol was introduced.  This
          information permits detailed analysis of the symbols used by a
          particular package and construction of an accurate dependency,
          but it requires the package maintainer to track more information
          about the shared library.
        </p>

        <p>
          A <file>shlibs</file> file, in contrast, only documents the last
          time the library ABI changed in any way.  It only provides
          information about the library as a whole, not individual
          symbols.  When a package is built using a shared library with
          only a <file>shlibs</file> file, the generated dependency will
          require a version of the shared library equal to or newer than
          the version of the last ABI change.  This generates
          unnecessarily restrictive dependencies compared
          to <file>symbols</file> files if none of the symbols used by the
          package have changed.  This, in turn, may make upgrades
          needlessly complex and unnecessarily restrict use of the package
          on systems with older versions of the shared libraries.
        </p>

        <p>
          <file>shlibs</file> files also only support a limited range of
          library SONAMEs, making it difficult to use <file>shlibs</file>
          files in some unusual corner cases.<footnote>
            A <file>shlibs</file> file represents an SONAME as a library
            name and version number, such as <tt>libfoo VERSION</tt>,
            instead of recording the actual SONAME.  If the SONAME doesn't
            match one of the two expected formats
            (<tt>libfoo-VERSION.so</tt> or <tt>libfoo.so.VERSION</tt>), it
            cannot be represented.
          </footnote>
        </p>

        <p>
          <file>symbols</file> files are therefore recommended for most
          shared library packages since they provide more accurate
          dependencies.  For most C libraries, the additional detail
          required by <file>symbols</file> files is not too difficult to
          maintain.  However, maintaining exhaustive symbols information
          for a C++ library can be quite onerous, so <file>shlibs</file>
          files may be more appropriate for most C++ libraries.  Libraries
          with a corresponding udeb must also provide
          a <file>shlibs</file> file, since the udeb infrastructure does
          not use <file>symbols</file> files.
        </p>

        <sect1 id="dpkg-shlibdeps">
          <heading>Generating dependencies on shared libraries</heading>

          <p>
            When a package that contains any shared libraries or compiled
            binaries is built, it must run <prgn>dpkg-shlibdeps</prgn> on
            each shared library and compiled binary to determine the
            libraries used and hence the dependencies needed by the
            package.<footnote>
              <prgn>dpkg-shlibdeps</prgn> will use a program
              like <prgn>objdump</prgn> or <prgn>readelf</prgn> to find
              the libraries and the symbols in those libraries directly
              needed by the binaries or shared libraries in the package.
            </footnote>
            To do this, put a call to <prgn>dpkg-shlibdeps</prgn> into
            your <file>debian/rules</file> file in the source package.
            List all of the compiled binaries, libraries, or loadable
            modules in your package.<footnote>
              The easiest way to call <prgn>dpkg-shlibdeps</prgn>
              correctly is to use a package helper framework such
              as <package>debhelper</package>.  If you are
              using <package>debhelper</package>,
              the <prgn>dh_shlibdeps</prgn> program will do this work for
              you.  It will also correctly handle multi-binary packages.
            </footnote>
            <prgn>dpkg-shlibdeps</prgn> will use the <file>symbols</file>
            or <file>shlibs</file> files installed by the shared libraries
            to generate dependency information.  The package must then
            provide a substitution variable into which the discovered
            dependency information can be placed.
          </p>

          <p>
            If you are creating a udeb for use in the Debian Installer,
            you will need to specify that <prgn>dpkg-shlibdeps</prgn>
            should use the dependency line of type <tt>udeb</tt> by adding
            the <tt>-tudeb</tt> option<footnote>
              <prgn>dh_shlibdeps</prgn> from the <tt>debhelper</tt> suite
              will automatically add this option if it knows it is
              processing a udeb.
            </footnote>. If there is no dependency line of
            type <tt>udeb</tt> in the <file>shlibs</file>
            file, <prgn>dpkg-shlibdeps</prgn> will fall back to the
            regular dependency line.
          </p>

          <p>
            <prgn>dpkg-shlibdeps</prgn> puts the dependency information
            into the <file>debian/substvars</file> file by default, which
            is then used by <prgn>dpkg-gencontrol</prgn>.  You will need
            to place a <tt>${shlibs:Depends}</tt> variable in
            the <tt>Depends</tt> field in the control file of every binary
            package built by this source package that contains compiled
            binaries, libraries, or loadable modules.  If you have
            multiple binary packages, you will need to
            call <prgn>dpkg-shlibdeps</prgn> on each one which contains
            compiled libraries or binaries.  For example, you could use
            the <tt>-T</tt> option to the <tt>dpkg</tt> utilities to
            specify a different <file>substvars</file> file for each
            binary package.<footnote>
              Again, <prgn>dh_shlibdeps</prgn>
              and <prgn>dh_gencontrol</prgn> will handle everything except
              the addition of the variable to the control file for you if
              you're using <package>debhelper</package>, including
              generating separate <file>substvars</file> files for each
              binary package and calling <prgn>dpkg-gencontrol</prgn> with
              the appropriate flags.
            </footnote>
          </p>

          <p>
            For more details on <prgn>dpkg-shlibdeps</prgn>,
            see <manref name="dpkg-shlibdeps" section="1">.
          </p>

          <p>
            We say that a binary <tt>foo</tt> <em>directly</em> uses a
            library <tt>libbar</tt> if it is explicitly linked with that
            library (that is, the library is listed in the
            ELF <tt>NEEDED</tt> attribute, caused by adding <tt>-lbar</tt>
            to the link line when the binary is created).  Other libraries
            that are needed by <tt>libbar</tt> are
            linked <em>indirectly</em> to <tt>foo</tt>, and the dynamic
            linker will load them automatically when it
            loads <tt>libbar</tt>.  A package should depend on the
            libraries it directly uses, but not the libraries it only uses
            indirectly.  The dependencies for the libraries used
            directly will automatically pull in the indirectly-used
            libraries.  <prgn>dpkg-shlibdeps</prgn> will handle this logic
            automatically, but package maintainers need to be aware of
            this distinction between directly and indirectly using a
            library if they have to override its results for some reason.
            <footnote>
              A good example of where this helps is the following.  We
              could update <tt>libimlib</tt> with a new version that
              supports a new revision of a graphics format called dgf (but
              retaining the same major version number) and depends on a
              new library package <package>libdgf4</package> instead of
              the older <package>libdgf3</package>.  If we
              used <prgn>ldd</prgn> to add dependencies for every library
              directly or indirectly linked with a binary, every package
              that uses <tt>libimlib</tt> would need to be recompiled so
              it would also depend on <package>libdgf4</package> in order
              to retire the older <package>libdgf3</package> package.
              Since dependencies are only added based on
              ELF <tt>NEEDED</tt> attribute, packages
              using <tt>libimlib</tt> can rely on <tt>libimlib</tt> itself
              having the dependency on an appropriate version
              of <tt>libdgf</tt> and do not need rebuilding.
            </footnote>
          </p>
        </sect1>

        <sect1 id="sharedlibs-updates">
          <heading>Shared library ABI changes</heading>

          <p>
            Maintaining a shared library package using
            either <file>symbols</file> or <file>shlibs</file> files
            requires being aware of the exposed ABI of the shared library
            and any changes to it.  Both <file>symbols</file>
            and <file>shlibs</file> files record every change to the ABI
            of the shared library; <file>symbols</file> files do so per
            public symbol, whereas <file>shlibs</file> files record only
            the last change for the entire library.
          </p>

          <p>
            There are two types of ABI changes: ones that are
            backward-compatible and ones that are not.  An ABI change is
            backward-compatible if any reasonable program or library that
            was linked with the previous version of the shared library
            will still work correctly with the new version of the shared
            library.<footnote>
              An example of an "unreasonable" program is one that uses
              library interfaces that are documented as internal and
              unsupported.  If the only programs or libraries affected by
              a change are "unreasonable" ones, other techniques, such as
              declaring <tt>Breaks</tt> relationships with affected
              packages or treating their usage of the library as bugs in
              those packages, may be appropriate instead of changing the
              SONAME.  However, the default approach is to change the
              SONAME for any change to the ABI that could break a program.
            </footnote>
            Adding new symbols to the shared library is a
            backward-compatible change.  Removing symbols from the shared
            library is not.  Changing the behavior of a symbol may or may
            not be backward-compatible depending on the change; for
            example, changing a function to accept a new enum constant not
            previously used by the library is generally
            backward-compatible, but changing the members of a struct that
            is passed into library functions is generally not unless the
            library takes special precautions to accept old versions of
            the data structure.
          </p>

          <p>
            ABI changes that are not backward-compatible normally require
            changing the <tt>SONAME</tt> of the library and therefore the
            shared library package name, which forces rebuilding all
            packages using that shared library to update their
            dependencies and allow them to use the new version of the
            shared library.  For more information,
            see <ref id="sharedlibs-runtime">.  The remainder of this
            section will deal with backward-compatible changes.
          </p>

          <p>
            Backward-compatible changes require either updating or
            recording the <var>minimal-version</var> for that symbol
            in <file>symbols</file> files or updating the version in
            the <var>dependencies</var> in <file>shlibs</file> files.  For
            more information on how to do this in the two formats, see
            <ref id="symbols"> and <ref id="shlibs">.  Below are general
            rules that apply to both files.
          </p>

          <p>
            The easy case is when a public symbol is added.  Simply add
            the version at which the symbol was introduced
            (for <file>symbols</file> files) or update the dependency
            version (for <file>shlibs</file>) files.  But special care
            should be taken to update dependency versions when the
            behavior of a public symbol changes.  This is easy to neglect,
            since there is no automated method of determining such
            changes, but failing to update versions in this case may
            result in binary packages with too-weak dependencies that will
            fail at runtime, possibly in ways that can cause security
            vulnerabilities.  If the package maintainer believes that a
            symbol behavior change may have occurred but isn't sure, it's
            safer to update the version rather than leave it unmodified.
            This may result in unnecessarily strict dependencies, but it
            ensures that packages whose dependencies are satisfied will
            work properly.
          </p>

          <p>
            A common example of when a change to the dependency version
            is required is a function that takes an enum or struct
            argument that controls what the function does.  For example:
            <example>
              enum library_op { OP_FOO, OP_BAR };
              int library_do_operation(enum library_op);
            </example>
            If a new operation, <tt>OP_BAZ</tt>, is added,
            the <var>minimal-version</var>
            of <tt>library_do_operation</tt> (for <file>symbols</file>
            files) or the version in the dependency for the shared library
            (for <file>shlibs</file> files) must be increased to the
            version at which <tt>OP_BAZ</tt> was introduced.  Otherwise, a
            binary built against the new version of the library (having
            detected at compile-time that the library
            supports <tt>OP_BAZ</tt>) may be installed with a shared
            library that doesn't support <tt>OP_BAZ</tt> and will fail at
            runtime when it tries to pass <tt>OP_BAZ</tt> into this
            function.
          </p>

          <p>
            Dependency versions in either <file>symbols</file>
            or <file>shlibs</file> files normally should not contain the
            Debian revision of the package, since the library behavior is
            normally fixed for a particular upstream version and any
            Debian packaging of that upstream version will have the same
            behavior.  In the rare case that the library behavior was
            changed in a particular Debian revision, appending <tt>~</tt>
            to the end of the version that includes the Debian revision is
            recommended, since this allows backports of the shared library
            package using the normal backport versioning convention to
            satisfy the dependency.
          </p>
        </sect1>

        <sect1 id="sharedlibs-symbols">
          <heading>The <tt>symbols</tt> system</heading>

          <p>
            In the following sections, we will first describe where the
            various <file>symbols</file> files are to be found, then
            the <file>symbols</file> file format, and finally how to
            create <file>symbols</file> files if your package contains a
            shared library.
          </p>

          <sect2 id="symbols-paths">
            <heading>The <file>symbols</file> files present on the
              system</heading>

            <p>
              <file>symbols</file> files for a shared library are normally
              provided by the shared library package as a control file,
              but there are several override paths that are checked first
              in case that information is wrong or missing.  The following
              list gives them in the order in which they are read
              by <prgn>dpkg-shlibdeps</prgn> The first one that contains
              the required information is used.
              <list>
                <item>
                  <p><file>debian/*/DEBIAN/symbols</file></p>

                  <p>
                    During the package build, if the package itself
                    contains shared libraries with <file>symbols</file>
                    files, they will be generated in these staging
                    directories by <prgn>dpkg-gensymbols</prgn>
                    (see <ref id="providing-symbols">).  <file>symbols</file>
                    files found in the build tree take precedence
                    over <file>symbols</file> files from other binary
                    packages.
                  </p>

                  <p>
                    These files must exist
                    before <prgn>dpkg-shlibdeps</prgn> is run or the
                    dependencies of binaries and libraries from a source
                    package on other libraries from that same source
                    package will not be correct.  In practice, this means
                    that <prgn>dpkg-gensymbols</prgn> must be run
                    before <prgn>dpkg-shlibdeps</prgn> during the package
                    build.<footnote>
                      An example may clarify.  Suppose the source
                      package <tt>foo</tt> generates two binary
                      packages, <tt>libfoo2</tt> and <tt>foo-runtime</tt>.
                      When building the binary packages, the contents of
                      the packages are staged in the
                      directories <file>debian/libfoo2</file>
                      and <file>debian/foo-runtime</file> respectively.
                      (<file>debian/tmp</file> could be used instead of
                      one of these.)  Since <tt>libfoo2</tt> provides
                      the <tt>libfoo</tt> shared library, it will contain
                      a <tt>symbols</tt> file, which will be installed
                      in <file>debian/libfoo2/DEBIAN/symbols</file>,
                      eventually to be included as a control file in that
                      package.  When <prgn>dpkg-shlibdeps</prgn> is run on
                      the
                      executable <file>debian/foo-runtime/usr/bin/foo-prog</file>,
                      it will examine
                      the <file>debian/libfoo2/DEBIAN/symbols</file> file
                      to determine whether <tt>foo-prog</tt>'s library
                      dependencies are satisfied by any of the libraries
                      provided by <tt>libfoo2</tt>.  Since those binaries
                      were linked against the just-built shared library as
                      part of the build process, the <file>symbols</file>
                      file for the newly-built <tt>libfoo2</tt> must take
                      precedence over a <file>symbols</file> file for any
                      other <tt>libfoo2</tt> package already installed on
                      the system.
                    </footnote>
                  </p>
                </item>

                <item>
                  <p>
                    <file>/etc/dpkg/symbols/<var>package</var>.symbols.<var>arch</var></file>
                    and <file>/etc/dpkg/symbols/<var>package</var>.symbols</file>
                  </p>

                  <p>
                    Per-system overrides of shared library dependencies.
                    These files normally do not exist.  They are
                    maintained by the local system administrator and must
                    not be created by any Debian package.
                  </p>
                </item>

                <item>
                  <p><file>symbols</file> control files for packages
                    installed on the system</p>

                  <p>
                    The <file>symbols</file> control files for all the
                    packages currently installed on the system are
                    searched last.  This will be the most common source of
                    shared library dependency information.  These are
                    normally found
                    in <file>/var/lib/dpkg/info/*.symbols</file>, but
                    packages should not rely on this and instead should
                    use <tt>dpkg-query --control-path <var>package</var>
                    symbols</tt> if for some reason these files need to be
                    examined.
                  </p>
                </item>
              </list>
            </p>

            <p>
              Be aware that if a <file>debian/shlibs.local</file> exists
              in the source package, it will override
              any <file>symbols</file> files.  This is the only case where
              a <file>shlibs</file> is used despite <file>symbols</file>
              files being present.  See <ref id="shlibs-paths">
              and <ref id="sharedlibs-shlibdeps"> for more information.
            </p>
          </sect2>

          <sect2 id="symbols">
            <heading>The <file>symbols</file> File Format</heading>

            <p>
              The following documents the format of
              the <file>symbols</file> control file as included in binary
              packages.  These files are built from
              template <file>symbols</file> files in the source package
              by <prgn>dpkg-gensymbols</prgn>.  The template files support
              a richer syntax that allows <prgn>dpkg-gensymbols</prgn> to
              do some of the tedious work involved in
              maintaining <file>symbols</file> files, such as handling C++
              symbols or optional symbols that may not exist on particular
              architectures.  When writing <file>symbols</file> files for
              a shared library package, refer
              to <manref name="dpkg-gensymbols" section="1"> for the
              richer syntax.
            </p>

            <p>
              A <file>symbols</file> may contain one or more entries, one
              for each shared library contained in the package
              corresponding to that <file>symbols</file>.  Each entry has
              the following format:
            </p>

            <p>
              <example>
                <var>library-soname</var> <var>main-dependency-template</var>
                [| <var>alternative-dependency-template</var>]
                [...]
                [* <var>field-name</var>: <var>field-value</var>]
                [...]
                <var>symbol</var> <var>minimal-version</var>[ <var>id-of-dependency-template</var> ]
              </example>
            </p>

            <p>
              To explain this format, we'll use the the <tt>zlib1g</tt>
              package as an example, which (at the time of writing)
              installs the shared
              library <file>/usr/lib/libz.so.1.2.3.4</file>.  Mandatory
              lines will be described first, followed by optional lines.
            </p>

            <p>
              <var>library-soname</var> must contain exactly the value of
              the ELF <tt>SONAME</tt> attribute of the shared library.  In
              our example, this is <tt>libz.so.1</tt>.<footnote>
                This can be determined by using the command
                <example compact="compact">
                  readelf -d /usr/lib/libz.so.1.2.3.4 | grep SONAME
                </example>
              </footnote>
            </p>

            <p>
              <var>main-dependency-template</var> has the same syntax as a
              dependency field in a binary package control file, except
              that the string <tt>#MINVER#</tt> is replaced by a version
              restriction like <tt>(>= <var>version</var>)</tt> or by
              nothing if an unversioned dependency is deemed sufficient.
              The version restriction will be based on which symbols from
              the shared library are referenced and the version at which
              they were introduced (see below).  In nearly all
              cases, <var>main-dependency-template</var> will
              be <tt><var>package</var> #MINVER#</tt>,
              where <var>package</var> is the name of the binary package
              containing the shared library.  This adds a simple,
              possibly-versioned dependency on the shared library package.
              In some rare cases, such as when multiple packages provide
              the same shared library ABI, the dependency template may
              need to be more complex.
            </p>

            <p>
              In our example, the first line of
              the <tt>zlib1g</tt> <file>symbols</file> file would be:
              <example compact="compact">
                libz.so.1 zlib1g #MINVER#
              </example>
            </p>

            <p>
              Each public symbol exported by the shared library must have
              a corresponding symbol line, indented by one
              space.  <var>symbol</var> is the exported symbol (which, for
              C++, means the mangled symbol) followed by <tt>@</tt> and
              the symbol version, or the string <tt>Base</tt> if there is
              no symbol version.  <var>minimal-version</var> is the most
              recent version of the shared library that changed the
              behavior of that symbol, whether by adding it, changing its
              function signature (the parameters, their types, or the
              return type), or changing its behavior in a way that is
              visible to a caller.
              <var>id-of-dependency-template</var> is an optional
              field that references
              an <var>alternative-dependency-template</var>; see below for
              a full description.
            </p>

            <p>
              For example, <tt>libz.so.1</tt> contains the
              symbols <tt>compress</tt>
              and <tt>compressBound</tt>.  <tt>compress</tt> has no symbol
              version and last changed its behavior in upstream
              version <tt>1:1.1.4</tt>.  <tt>compressBound</tt> has the
              symbol version <tt>ZLIB_1.2.0</tt>, was introduced in
              upstream version <tt>1:1.2.0</tt>, and has not changed its
              behavior.  Its <file>symbols</file> file therefore contains
              the lines:
              <example compact="compact">
                compress@Base 1:1.1.4
                compressBound@ZLIB_1.2.0 1:1.2.0
              </example>
              Packages using only <tt>compress</tt> would then get a
              dependency on <tt>zlib1g (>= 1:1.1.4)</tt>, but packages
              using <tt>compressBound</tt> would get a dependency
              on <tt>zlib1g (>= 1:1.2.0)</tt>.
            </p>

            <p>
              One or more <var>alternative-dependency-template</var> lines
              may be provided.  These are used in cases where some symbols
              in the shared library should use one dependency template
              while others should use a different template.  The
              alternative dependency templates are used only if a symbol
              line contains the <var>id-of-dependency-template</var>
              field.  The first alternative dependency template is
              numbered 1, the second 2, and so forth.<footnote>
                An example of where this may be needed is with a library
                that implements the libGL interface.  All GL
                implementations provide the same set of base interfaces,
                and then may provide some additional interfaces only used
                by programs that require that specific GL implementation.
                So, for example, libgl1-mesa-glx may use the
                following <file>symbols</file> file:
                <example>
                  libGL.so.1 libgl1
                  | libgl1-mesa-glx #MINVER#
                  publicGlSymbol@Base 6.3-1
                  [...]
                  implementationSpecificSymbol@Base 6.5.2-7 1
                  [...]
                </example>
                Binaries or shared libraries using
                only <tt>publicGlSymbol</tt> would depend only
                on <tt>libgl1</tt> (which may be provided by multiple
                packages), but ones
                using <tt>implementationSpecificSymbol</tt> would get a
                dependency on <tt>libgl1-mesa-glx (>= 6.5.2-7)</tt>
              </footnote>
            </p>

            <p>
              Finally, the entry for the library may contain one or more
              metadata fields.  Currently, the only
              supported <var>field-name</var>
              is <tt>Build-Depends-Package</tt>, whose value lists
              the <qref id="sharedlibs-dev">library development
              package</qref> on which packages using this shared library
              declare a build dependency.  If this field is
              present, <prgn>dpkg-shlibdeps</prgn> uses it to ensure that
              the resulting binary package dependency on the shared
              library is at least as strict as the source package
              dependency on the shared library development
              package.<footnote>
                This field should normally not be necessary, since if the
                behavior of any symbol has changed, the corresponding
                symbol <var>minimal-version</var> should have been
                increased.  But including it makes the <tt>symbols</tt>
                system more robust by tightening the dependency in cases
                where the package using the shared library specifically
                requires at least a particular version of the shared
                library development package for some reason.
              </footnote>
              For our example, the <tt>zlib1g</tt> <file>symbols</file>
              file would contain:
              <example compact="compact">
                * Build-Depends-Package: zlib1g-dev
              </example>
            </p>

            <p>
              Also see <manref name="deb-symbols" section="5">.
            </p>
          </sect2>

          <sect2 id="providing-symbols">
            <heading>Providing a <file>symbols</file> file</heading>

            <p>
              If your package provides a shared library, you should
              arrange to include a <file>symbols</file> control file
              following the format described above in that package.  You
              must include either a <file>symbols</file> control file or
              a <file>shlibs</file> control file.
            </p>

            <p>
              Normally, this is done by creating a <file>symbols</file> in
              the source package
              named <file>debian/<var>package</var>.symbols</file>
              or <file>debian/symbols</file>, possibly
              with <file>.<var>arch</var></file> appended if the symbols
              information varies by architecture.  This file may use the
              extended syntax documented in <manref name="dpkg-gensymbols"
              section="1">.  Then, call <prgn>dpkg-gensymbols</prgn> as
              part of the package build process.  It will
              create <file>symbols</file> files in the package staging
              area based on the binaries and libraries in the package
              staging area and the <file>symbols</file> files in the
              source package.<footnote>
                If you are
                using <tt>debhelper</tt>, <prgn>dh_makeshlibs</prgn> will
                take care of calling either <prgn>dpkg-gensymbols</prgn>
                or generating a <file>shlibs</file> file as appropriate.
              </footnote>
            </p>

            <p>
              Packages that provide <file>symbols</file> files must keep
              them up-to-date to ensure correct dependencies in packages
              that use the shared libraries.  This means updating
              the <file>symbols</file> file whenever a new public symbol
              is added, changing the <var>minimal-version</var> field
              whenever a symbol changes behavior or signature in a
              backward-compatible way (see <ref id="sharedlibs-updates">),
              and changing the <var>library-soname</var>
              and <var>main-dependency-template</var>, and probably all of
              the <var>minimal-version</var> fields, when the library
              changes <tt>SONAME</tt>.  Removing a public symbol from
              the <file>symbols</file> file because it's no longer
              provided by the library normally requires changing
              the <tt>SONAME</tt> of the library.
              See <ref id="sharedlibs-runtime"> for more information
              on <tt>SONAME</tt>s.
            </p>
          </sect2>
        </sect1>

        <sect1 id="sharedlibs-shlibdeps">
          <heading>The <tt>shlibs</tt> system</heading>

          <p>
            The <tt>shlibs</tt> system is a simpler alternative to
            the <tt>symbols</tt> system for declaring dependencies for
            shared libraries.  It may be more appropriate for C++
            libraries and other cases where tracking individual symbols is
            too difficult.  It predated the <tt>symbols</tt> system and is
            therefore frequently seen in older packages.  It is also
            required for udebs, which do not support <tt>symbols</tt>.
          </p>

          <p>
            In the following sections, we will first describe where the
            various <file>shlibs</file> files are to be found, then how to
            use <prgn>dpkg-shlibdeps</prgn>, and finally
            the <file>shlibs</file> file format and how to create them.
          </p>

          <sect2 id="shlibs-paths">
            <heading>The <file>shlibs</file> files present on the
              system</heading>

            <p>
              There are several places where <tt>shlibs</tt> files are
              found.  The following list gives them in the order in which
              they are read by <prgn>dpkg-shlibdeps</prgn>.  (The first
              one which gives the required information is used.)
              <list>
                <item>
                  <p><file>debian/shlibs.local</file></p>

                  <p>
                    This lists overrides for this package.  This file
                    should normally not be used, but may be needed
                    temporarily in unusual situations to work around bugs
                    in other packages, or in unusual cases where the
                    normally declared dependency information in the
                    installed <file>shlibs</file> file for a library
                    cannot be used.  This file overrides information
                    obtained from any other source.
                  </p>
                </item>

                <item>
                  <p><file>/etc/dpkg/shlibs.override</file></p>

                  <p>
                    This lists global overrides.  This list is normally
                    empty.  It is maintained by the local system
                    administrator.
                  </p>
                </item>

                <item>
                  <p><file>DEBIAN/shlibs</file> files in the "build
                    directory"</p>

                  <p>
                    These files are generated as part of the package build
                    process and staged for inclusion as control files in
                    the binary packages being built.  They provide details
                    of any shared libraries included in the same package.
                  </p>
                </item>

                <item>
                  <p><file>shlibs</file> control files for packages
                    installed on the system</p>

                  <p>
                    The <file>shlibs</file> control files for all the
                    packages currently installed on the system.  These are
                    normally found
                    in <file>/var/lib/dpkg/info/*.shlibs</file>, but
                    packages should not rely on this and instead should
                    use <tt>dpkg-query --control-path <var>package</var>
                    shlibs</tt> if for some reason these files need to be
                    examined.
                  </p>
                </item>

                <item>
                  <p><file>/etc/dpkg/shlibs.default</file></p>

                  <p>
                    This file lists any shared libraries whose packages
                    have failed to provide correct <file>shlibs</file>
                    files.  It was used when the <file>shlibs</file> setup
                    was first introduced, but it is now normally empty.
                    It is maintained by the <tt>dpkg</tt> maintainer.
                  </p>
                </item>
              </list>
            </p>

            <p>
              If a <file>symbols</file> file for a shared library package
              is available, <prgn>dpkg-shlibdeps</prgn> will always use it
              in preference to a <file>shlibs</file>, with the exception
              of <file>debian/shlibs.local</file>.  The latter overrides
              any other <file>shlibs</file> or <file>symbols</file> files.
            </p>
          </sect2>

          <sect2 id="shlibs">
            <heading>The <file>shlibs</file> File Format</heading>

            <p>
              Each <file>shlibs</file> file has the same format.  Lines
              beginning with <tt>#</tt> are considered to be comments and
              are ignored.  Each line is of the form:
              <example compact="compact">
                [<var>type</var>: ]<var>library-name</var> <var>soname-version</var> <var>dependencies ...</var>
              </example>
            </p>

            <p>
              We will explain this by reference to the example of the
              <tt>zlib1g</tt> package, which (at the time of writing)
              installs the shared
              library <file>/usr/lib/libz.so.1.2.3.4</file>.
            </p>

            <p>
              <var>type</var> is an optional element that indicates the
              type of package for which the line is valid. The only type
              currently in use is <tt>udeb</tt>.  The colon and space
              after the type are required.
            </p>

            <p>
              <var>library-name</var> is the name of the shared library,
              in this case <tt>libz</tt>.  (This must match the name part
              of the soname, see below.)
            </p>

            <p>
              <var>soname-version</var> is the version part of the
              ELF <tt>SONAME</tt> attribute of the library, determined the
              same way that the <var>soversion</var> component of the
              recommended shared library package name is determined.
              See <ref id="sharedlibs-runtime"> for the details.
            </p>

            <p>
              <var>dependencies</var> has the same syntax as a dependency
              field in a binary package control file.  It should give
              details of which packages are required to satisfy a binary
              built against the version of the library contained in the
              package.  See <ref id="depsyntax"> for details on the
              syntax, and <ref id="sharedlibs-updates"> for details on how
              to maintain the dependency version constraint.
            </p>

            <p>
              In our example, if the last change to the <tt>zlib1g</tt>
              package that could change behavior for a client of that
              library was in version <tt>1:1.2.3.3.dfsg-1</tt>, then
              the <tt>shlibs</tt> entry for this library could say:
              <example compact="compact">
                libz 1 zlib1g (>= 1:1.2.3.3.dfsg)
              </example>
              This version restriction must be new enough that any binary
              built against the current version of the library will work
              with any version of the shared library that satisfies that
              dependency.
            </p>

            <p>
              As zlib1g also provides a udeb containing the shared
              library, there would also be a second line:
              <example compact="compact">
                udeb: libz 1 zlib1g-udeb (>= 1:1.2.3.3.dfsg)
              </example>
            </p>
          </sect2>

          <sect2>
            <heading>Providing a <file>shlibs</file> file</heading>

            <p>
              To provide a <file>shlibs</file> file for a shared library
              binary package, create a <file>shlibs</file> file following
              the format described above and place it in
              the <file>DEBIAN</file> directory for that package during
              the build.  It will then be included as a control file for
              that package<footnote>
                This is what <prgn>dh_makeshlibs</prgn> in
                the <package>debhelper</package> suite does. If your
                package also has a udeb that provides a shared
                library, <prgn>dh_makeshlibs</prgn> can automatically
                generate the <tt>udeb:</tt> lines if you specify the name
                of the udeb with the <tt>--add-udeb</tt> option.
              </footnote>.
            </p>

            <p>
              Since <prgn>dpkg-shlibdeps</prgn> reads
              the <file>DEBIAN/shlibs</file> files in all of the binary
              packages being built from this source package, all of
              the <file>DEBIAN/shlibs</file> files should be installed
              before <prgn>dpkg-shlibdeps</prgn> is called on any of the
              binary packages.
            </p>
          </sect2>
        </sect1>
      </sect>
    </chapt>


    <chapt id="opersys"><heading>The Operating System</heading>

      <sect>
        <heading>File system hierarchy</heading>


        <sect1 id="fhs">
          <heading>File System Structure</heading>

          <p>
            The location of all files and directories must comply with the
            Filesystem Hierarchy Standard (FHS), version 2.3, with the
            exceptions noted below, and except where doing so would
            violate other terms of Debian Policy.  The following
            exceptions to the FHS apply:

            <enumlist>
              <item>
                <p>
                The FHS requirement that architecture-independent
                application-specific static files be located in
                <file>/usr/share</file> is relaxed to a suggestion.

                In particular, a subdirectory of <file>/usr/lib</file> may
                be used by a package (or a collection of packages) to hold a
                mixture of architecture-independent and
                architecture-dependent files. However, when a directory is
                entirely composed of architecture-independent files, it
                should be located in <file>/usr/share</file>.
                </p>
              </item>
              <item>
                <p>
                  The optional rules related to user specific
                  configuration files for applications are stored in
                  the user's home directory are relaxed.  It is
                  recommended that such files start with the
                  '<tt>.</tt>' character (a "dot file"), and if an
                  application needs to create more than one dot file
                  then the preferred placement is in a subdirectory
                  with a name starting with a '.' character, (a "dot
                  directory"). In this case it is recommended the
                  configuration files not start with the '.'
                  character.
                </p>
              </item>
              <item>
                <p>
                  The requirement for amd64 to use <file>/lib64</file>
                  for 64 bit binaries is removed.
                </p>
              </item>
              <item>
                <p>
                  The requirement for object files, internal binaries, and
                  libraries, including <file>libc.so.*</file>, to be located
                  directly under <file>/lib{,32}</file> and
                  <file>/usr/lib{,32}</file> is amended, permitting files
                  to instead be installed to
                  <file>/lib/<var>triplet</var></file> and
                  <file>/usr/lib/<var>triplet</var></file>, where
                  <tt><var>triplet</var></tt> is the value returned by
                  <tt>dpkg-architecture -qDEB_HOST_MULTIARCH</tt> for the
                  architecture of the package.  Packages may <em>not</em>
                  install files to any <var>triplet</var> path other
                  than the one matching the architecture of that package;
                  for instance, an <tt>Architecture: amd64</tt> package
                  containing 32-bit x86 libraries may not install these
                  libraries to <file>/usr/lib/i386-linux-gnu</file>.
                  <footnote>
                    This is necessary in order to reserve the directories for
                    use in cross-installation of library packages from other
                    architectures, as part of <tt>multiarch</tt>.
                  </footnote>
                </p>
                <p>
                  The requirement for C and C++ headers files to be
                  accessible through the search path
                  <file>/usr/include/</file> is amended, permitting files to
                  be accessible through the search path
                  <file>/usr/include/<var>triplet</var></file> where
                  <tt><var>triplet</var></tt> is as above.  <footnote>
                    This is necessary for architecture-dependant headers
                    file to coexist in a <tt>multiarch</tt> setup.
                  </footnote>
                </p>
                <p>
                  Applications may also use a single subdirectory under
                  <file>/usr/lib/<var>triplet</var></file>.
                </p>
                <p>
                  The execution time linker/loader, ld*, must still be made
                  available in the existing location under /lib or /lib64
                  since this is part of the ELF ABI for the architecture.
                </p>
              </item>
              <item>
                <p>
                  The requirement that
                  <file>/usr/local/share/man</file> be "synonymous"
                  with <file>/usr/local/man</file> is relaxed to a
                  recommendation</p>
              </item>
              <item>
                <p>
                  The requirement that windowmanagers with a single
                  configuration file call it <file>system.*wmrc</file>
                  is removed, as is the restriction that the window
                  manager subdirectory be named identically to the
                  window manager name itself.
                </p>
              </item>
              <item>
                <p>
                  The requirement that boot manager configuration
                  files live in <file>/etc</file>, or at least are
                  symlinked there, is relaxed to a recommendation.
                </p>
              </item>
              <item>
                <p>
                  The additional directory <file>/run</file> in the root
                  file system is allowed.  <file>/run</file>
                  replaces <file>/var/run</file>, and the
                  subdirectory <file>/run/lock</file>
                  replaces <file>/var/lock</file>, with
                  the <file>/var</file> directories replaced by symlinks
                  for backwards compatibility.  <file>/run</file>
                  and <file>/run/lock</file> must follow all of the
                  requirements in the FHS for <file>/var/run</file>
                  and <file>/var/lock</file>, respectively, such as file
                  naming conventions, file format requirements, or the
                  requirement that files be cleared during the boot
                  process.  Files and directories residing
                  in <file>/run</file> should be stored on a temporary
                  file system.
                </p>
                <p>
                  Packages must not assume the <file>/run</file>
                  directory exists or is usable without a dependency
                  on <tt>initscripts (>= 2.88dsf-13.3)</tt> until the
                  stable release of Debian supports <file>/run</file>.
                </p>
              </item>
              <item>
                <p>
                  The <file>/sys</file> directory in the root filesystem is
                  additionally allowed. <footnote>This directory is used as
                    mount point to mount virtual filesystems to get access to
                    kernel information.</footnote>
                </p>
              </item>
              <item>
                <p>
                  The requirement for <file>/usr/local/lib&lt;qual&gt;</file>
                  to exist if <file>/lib&lt;qual&gt;</file> or
                  <file>/usr/lib&lt;qual&gt;</file> exists (where 
                  <file>lib&lt;qual&gt;</file> is a variant of
                  <file>lib</file> such as <file>lib32</file> or
                  <file>lib64</file>) is removed.
                  </p>
              </item>
              <item>
                <p>
                  On GNU/Hurd systems, the following additional
                  directories are allowed in the root
                  filesystem: <file>/hurd</file>
                  and <file>/servers</file>.<footnote>
                    These directories are used to store translators and as
                    a set of standard names for mount points,
                    respectively.
                  </footnote>
                </p>
              </item>
            </enumlist>
          </p>

          <p>
            The version of this document referred here can be
            found in the <tt>debian-policy</tt> package or on <url
            id="http://www.debian.org/doc/packaging-manuals/fhs/"
              name="FHS (Debian copy)"> alongside this manual (or, if
            you have the <package>debian-policy</package> installed,
            you can try <url
              id="file:///usr/share/doc/debian-policy/fhs/" name="FHS
              (local copy)">). The
            latest version, which may be a more recent version, may
            be found on
            <url id="http://www.pathname.com/fhs/" name="FHS (upstream)">.
            Specific questions about following the standard may be
            asked on the <tt>debian-devel</tt> mailing list, or
            referred to the FHS mailing list (see the
            <url id="http://www.pathname.com/fhs/" name="FHS web site"> for
            more information).
          </p>
        </sect1>

        <sect1>
          <heading>Site-specific programs</heading>

          <p>
            As mandated by the FHS, packages must not place any
            files in <file>/usr/local</file>, either by putting them in
            the file system archive to be unpacked by <prgn>dpkg</prgn>
            or by manipulating them in their maintainer scripts.
          </p>

          <p>
            However, the package may create empty directories below
            <file>/usr/local</file> so that the system administrator knows
            where to place site-specific files.  These are not
            directories <em>in</em> <file>/usr/local</file>, but are
            children of directories in <file>/usr/local</file>.  These
            directories (<file>/usr/local/*/dir/</file>)
            should be removed on package removal if they are
            empty.
          </p>

          <p>
            Note that this applies only to
            directories <em>below</em> <file>/usr/local</file>,
            not <em>in</em> <file>/usr/local</file>.  Packages must
            not create sub-directories in the
            directory <file>/usr/local</file> itself, except those
            listed in FHS, section 4.5.  However, you may create
            directories below them as you wish. You must not remove
            any of the directories listed in 4.5, even if you created
            them.
          </p>

          <p>
            Since <file>/usr/local</file> can be mounted read-only from a
            remote server, these directories must be created and
            removed by the <prgn>postinst</prgn> and <prgn>prerm</prgn>
            maintainer scripts and not be included in the
            <file>.deb</file> archive.  These scripts must not fail if
            either of these operations fail.
          </p>

          <p>
            For example, the <tt>emacsen-common</tt> package could
            contain something like
            <example compact="compact">
if [ ! -e /usr/local/share/emacs ]; then
  if mkdir /usr/local/share/emacs 2>/dev/null; then
    if chown root:staff /usr/local/share/emacs; then
      chmod 2775 /usr/local/share/emacs || true
    fi
  fi
fi
            </example>
            in its <prgn>postinst</prgn> script, and
            <example compact="compact">
rmdir /usr/local/share/emacs/site-lisp 2>/dev/null || true
rmdir /usr/local/share/emacs 2>/dev/null || true
            </example>
            in the <prgn>prerm</prgn> script.  (Note that this form is
            used to ensure that if the script is interrupted, the
            directory <file>/usr/local/share/emacs</file> will still be
            removed.)
          </p>

          <p>
            If you do create a directory in <file>/usr/local</file> for
            local additions to a package, you should ensure that
            settings in <file>/usr/local</file> take precedence over the
            equivalents in <file>/usr</file>.
          </p>

          <p>
            However, because <file>/usr/local</file> and its contents are
            for exclusive use of the local administrator, a package
            must not rely on the presence or absence of files or
            directories in <file>/usr/local</file> for normal operation.
          </p>

          <p>
            The <file>/usr/local</file> directory itself and all the
            subdirectories created by the package should (by default) have
            permissions 2775 (group-writable and set-group-id) and be
            owned by <tt>root:staff</tt>.
          </p>
        </sect1>

        <sect1>
          <heading>The system-wide mail directory</heading>
          <p>
            The system-wide mail directory
            is <file>/var/mail</file>. This directory is part of the
            base system and should not be owned by any particular mail
            agents.  The use of the old
            location <file>/var/spool/mail</file> is deprecated, even
            though the spool may still be physically located there.
          </p>
        </sect1>

        <sect1 id="fhs-run">
          <heading><file>/run</file> and <file>/run/lock</file></heading>

          <p>
            The directory <file>/run</file> is cleared at boot, normally
            by being a mount point for a temporary file system.  Packages
            therefore must not assume that any files or directories
            under <file>/run</file> other than <file>/run/lock</file>
            exist unless the package has arranged to create those files or
            directories since the last reboot.  Normally, this is done by
            the package via an init script.  See <ref id="writing-init">
            for more information.
          </p>

          <p>
            Packages must not include files or directories
            under <file>/run</file>, or under the
            older <file>/var/run</file> and <file>/var/lock</file> paths.
            The latter paths will normally be symlinks or other
            redirections to <file>/run</file> for backwards compatibility.
          </p>
        </sect1>
      </sect>

      <sect>
        <heading>Users and groups</heading>

        <sect1>
          <heading>Introduction</heading>
          <p>
            The Debian system can be configured to use either plain or
            shadow passwords.
          </p>

          <p>
            Some user ids (UIDs) and group ids (GIDs) are reserved
            globally for use by certain packages.  Because some
            packages need to include files which are owned by these
            users or groups, or need the ids compiled into binaries,
            these ids must be used on any Debian system only for the
            purpose for which they are allocated. This is a serious
            restriction, and we should avoid getting in the way of
            local administration policies. In particular, many sites
            allocate users and/or local system groups starting at 100.
          </p>

          <p>
            Apart from this we should have dynamically allocated ids,
            which should by default be arranged in some sensible
            order, but the behavior should be configurable.
          </p>

          <p>
            Packages other than <tt>base-passwd</tt> must not modify
            <file>/etc/passwd</file>, <file>/etc/shadow</file>,
            <file>/etc/group</file> or <file>/etc/gshadow</file>.
          </p>
        </sect1>

        <sect1>
          <heading>UID and GID classes</heading>
          <p>
            The UID and GID numbers are divided into classes as
            follows:
            <taglist>
              <tag>0-99:</tag>
              <item>
                <p>
                  Globally allocated by the Debian project, the same
                  on every Debian system.  These ids will appear in
                  the <file>passwd</file> and <file>group</file> files of all
                  Debian systems, new ids in this range being added
                  automatically as the <tt>base-passwd</tt> package is
                  updated.
                </p>

                <p>
                  Packages which need a single statically allocated
                  uid or gid should use one of these; their
                  maintainers should ask the <tt>base-passwd</tt>
                  maintainer for ids.
                </p>
              </item>

              <tag>100-999:</tag>
              <item>
                <p>
                  Dynamically allocated system users and groups.
                  Packages which need a user or group, but can have
                  this user or group allocated dynamically and
                  differently on each system, should use <tt>adduser
                  --system</tt> to create the group and/or user.
                  <prgn>adduser</prgn> will check for the existence of
                  the user or group, and if necessary choose an unused
                  id based on the ranges specified in
                  <file>adduser.conf</file>.
                </p>
              </item>

              <tag>1000-59999:</tag>
              <item>
                <p>
                  Dynamically allocated user accounts.  By default
                  <prgn>adduser</prgn> will choose UIDs and GIDs for
                  user accounts in this range, though
                  <file>adduser.conf</file> may be used to modify this
                  behavior.
                </p>
              </item>

              <tag>60000-64999:</tag>
              <item>
                <p>
                  Globally allocated by the Debian project, but only
                  created on demand. The ids are allocated centrally
                  and statically, but the actual accounts are only
                  created on users' systems on demand.
                </p>

                <p>
                  These ids are for packages which are obscure or
                  which require many statically-allocated ids.  These
                  packages should check for and create the accounts in
                  <file>/etc/passwd</file> or <file>/etc/group</file> (using
                  <prgn>adduser</prgn> if it has this facility) if
                  necessary.  Packages which are likely to require
                  further allocations should have a "hole" left after
                  them in the allocation, to give them room to
                  grow.
                </p>
              </item>

              <tag>65000-65533:</tag>
              <item>
                <p>Reserved.</p>
              </item>

              <tag>65534:</tag>
              <item>
                <p>
                  User <tt>nobody</tt>. The corresponding gid refers
                  to the group <tt>nogroup</tt>.
                </p>
              </item>

              <tag>65535:</tag>
              <item>
                <p>
                  <tt>(uid_t)(-1) == (gid_t)(-1)</tt> <em>must
                  not</em> be used, because it is the error return
                  sentinel value.
                </p>
              </item>
            </taglist>
          </p>
        </sect1>
      </sect>

      <sect id="sysvinit">
        <heading>System run levels and <file>init.d</file> scripts</heading>

        <sect1 id="/etc/init.d">
          <heading>Introduction</heading>

          <p>
            The <file>/etc/init.d</file> directory contains the scripts
            executed by <prgn>init</prgn> at boot time and when the
            init state (or "runlevel") is changed (see <manref
            name="init" section="8">).
          </p>

          <p>
            There are at least two different, yet functionally
            equivalent, ways of handling these scripts.  For the sake
            of simplicity, this document describes only the symbolic
            link method. However, it must not be assumed by maintainer
            scripts that this method is being used, and any automated
            manipulation of the various runlevel behaviors by
            maintainer scripts must be performed using
            <prgn>update-rc.d</prgn> as described below and not by
            manually installing or removing symlinks.  For information
            on the implementation details of the other method,
            implemented in the <tt>file-rc</tt> package, please refer
            to the documentation of that package.
          </p>

          <p>
            These scripts are referenced by symbolic links in the
            <file>/etc/rc<var>n</var>.d</file> directories.  When changing
            runlevels, <prgn>init</prgn> looks in the directory
            <file>/etc/rc<var>n</var>.d</file> for the scripts it should
            execute, where <tt><var>n</var></tt> is the runlevel that
            is being changed to, or <tt>S</tt> for the boot-up
            scripts.
          </p>

          <p>
            The names of the links all have the form
            <file>S<var>mm</var><var>script</var></file> or
            <file>K<var>mm</var><var>script</var></file> where
            <var>mm</var> is a two-digit number and <var>script</var>
            is the name of the script (this should be the same as the
            name of the actual script in <file>/etc/init.d</file>).
          </p>

          <p>
            When <prgn>init</prgn> changes runlevel first the targets
            of the links whose names start with a <tt>K</tt> are
            executed, each with the single argument <tt>stop</tt>,
            followed by the scripts prefixed with an <tt>S</tt>, each
            with the single argument <tt>start</tt>.  (The links are
            those in the <file>/etc/rc<var>n</var>.d</file> directory
            corresponding to the new runlevel.)  The <tt>K</tt> links
            are responsible for killing services and the <tt>S</tt>
            link for starting services upon entering the runlevel.
          </p>

          <p>
            For example, if we are changing from runlevel 2 to
            runlevel 3, init will first execute all of the <tt>K</tt>
            prefixed scripts it finds in <file>/etc/rc3.d</file>, and then
            all of the <tt>S</tt> prefixed scripts in that directory.
            The links starting with <tt>K</tt> will cause the
            referred-to file to be executed with an argument of
            <tt>stop</tt>, and the <tt>S</tt> links with an argument
            of <tt>start</tt>.
          </p>

          <p>
            The two-digit number <var>mm</var> is used to determine
            the order in which to run the scripts: low-numbered links
            have their scripts run first.  For example, the
            <tt>K20</tt> scripts will be executed before the
            <tt>K30</tt> scripts.  This is used when a certain service
            must be started before another.  For example, the name
            server <prgn>bind</prgn> might need to be started before
            the news server <prgn>inn</prgn> so that <prgn>inn</prgn>
            can set up its access lists.  In this case, the script
            that starts <prgn>bind</prgn> would have a lower number
            than the script that starts <prgn>inn</prgn> so that it
            runs first:
            <example compact="compact">
/etc/rc2.d/S17bind
/etc/rc2.d/S70inn
            </example>
          </p>

          <p>
            The two runlevels 0 (halt) and 6 (reboot) are slightly
            different.  In these runlevels, the links with an
            <tt>S</tt> prefix are still called after those with a
            <tt>K</tt> prefix, but they too are called with the single
            argument <tt>stop</tt>.
          </p>
        </sect1>

        <sect1 id="writing-init">
          <heading>Writing the scripts</heading>

          <p>
            Packages that include daemons for system services should
            place scripts in <file>/etc/init.d</file> to start or stop
            services at boot time or during a change of runlevel.
            These scripts should be named
            <file>/etc/init.d/<var>package</var></file>, and they should
            accept one argument, saying what to do:

            <taglist>
              <tag><tt>start</tt></tag>
              <item>start the service,</item>

              <tag><tt>stop</tt></tag>
              <item>stop the service,</item>

              <tag><tt>restart</tt></tag>
              <item>stop and restart the service if it's already running,
                  otherwise start the service</item>

              <tag><tt>reload</tt></tag>
              <item><p>cause the configuration of the service to be
                  reloaded without actually stopping and restarting
                  the service,</item>

              <tag><tt>force-reload</tt></tag>
              <item>cause the configuration to be reloaded if the
                  service supports this, otherwise restart the
                  service.</item>
            </taglist>

            The <tt>start</tt>, <tt>stop</tt>, <tt>restart</tt>, and
            <tt>force-reload</tt> options should be supported by all
            scripts in <file>/etc/init.d</file>, the <tt>reload</tt>
            option is optional.
          </p>

          <p>
            The <file>init.d</file> scripts must ensure that they will
            behave sensibly (i.e., returning success and not starting
            multiple copies of a service) if invoked with <tt>start</tt>
            when the service is already running, or with <tt>stop</tt>
            when it isn't, and that they don't kill unfortunately-named
            user processes.  The best way to achieve this is usually to
            use <prgn>start-stop-daemon</prgn> with the <tt>--oknodo</tt>
            option.
          </p>

          <p>
            Be careful of using <tt>set -e</tt> in <file>init.d</file>
            scripts.  Writing correct <file>init.d</file> scripts requires
            accepting various error exit statuses when daemons are already
            running or already stopped without aborting
            the <file>init.d</file> script, and common <file>init.d</file>
            function libraries are not safe to call with <tt>set -e</tt>
            in effect<footnote>
              <tt>/lib/lsb/init-functions</tt>, which assists in writing
              LSB-compliant init scripts, may fail if <tt>set -e</tt> is
              in effect and echoing status messages to the console fails,
              for example.
            </footnote>.  For <tt>init.d</tt> scripts, it's often easier
            to not use <tt>set -e</tt> and instead check the result of
            each command separately.
          </p>

          <p>
            If a service reloads its configuration automatically (as
            in the case of <prgn>cron</prgn>, for example), the
            <tt>reload</tt> option of the <file>init.d</file> script
            should behave as if the configuration has been reloaded
            successfully.
          </p>

          <p>
            The <file>/etc/init.d</file> scripts must be treated as
            configuration files, either (if they are present in the
            package, that is, in the .deb file) by marking them as
            <tt>conffile</tt>s, or, (if they do not exist in the .deb)
            by managing them correctly in the maintainer scripts (see
            <ref id="config-files">).  This is important since we want
            to give the local system administrator the chance to adapt
            the scripts to the local system, e.g., to disable a
            service without de-installing the package, or to specify
            some special command line options when starting a service,
            while making sure their changes aren't lost during the next
            package upgrade.
          </p>

          <p>
            These scripts should not fail obscurely when the
            configuration files remain but the package has been
            removed, as configuration files remain on the system after
            the package has been removed.  Only when <prgn>dpkg</prgn>
            is executed with the <tt>--purge</tt> option will
            configuration files be removed.  In particular, as the
            <file>/etc/init.d/<var>package</var></file> script itself is
            usually a <tt>conffile</tt>, it will remain on the system
            if the package is removed but not purged.  Therefore, you
            should include a <tt>test</tt> statement at the top of the
            script, like this:
            <example compact="compact">
test -f <var>program-executed-later-in-script</var> || exit 0
            </example>
          </p>

          <p>
            Often there are some variables in the <file>init.d</file>
            scripts whose values control the behavior of the scripts,
            and which a system administrator is likely to want to
            change.  As the scripts themselves are frequently
            <tt>conffile</tt>s, modifying them requires that the
            administrator merge in their changes each time the package
            is upgraded and the <tt>conffile</tt> changes.  To ease
            the burden on the system administrator, such configurable
            values should not be placed directly in the script.
            Instead, they should be placed in a file in
            <file>/etc/default</file>, which typically will have the same
            base name as the <file>init.d</file> script.  This extra file
            should be sourced by the script when the script runs.  It
            must contain only variable settings and comments in SUSv3
            <prgn>sh</prgn> format.  It may either be a
            <tt>conffile</tt> or a configuration file maintained by
            the package maintainer scripts.  See <ref id="config-files">
            for more details.
          </p>

          <p>
            To ensure that vital configurable values are always
            available, the <file>init.d</file> script should set default
            values for each of the shell variables it uses, either
            before sourcing the <file>/etc/default/</file> file or
            afterwards using something like the <tt>:
            ${VAR:=default}</tt> syntax.  Also, the <file>init.d</file>
            script must behave sensibly and not fail if the
            <file>/etc/default</file> file is deleted.
          </p>

          <p>
            Files and directories under <file>/run</file>, including ones
            referred to via the compatibility paths <file>/var/run</file>
            and <file>/var/lock</file>, are normally stored on a temporary
            filesystem and are normally not persistent across a reboot.
            The <file>init.d</file> scripts must handle this correctly.
            This will typically mean creating any required subdirectories
            dynamically when the <file>init.d</file> script is run.
            See <ref id="fhs-run"> for more information.
          </p>
        </sect1>

        <sect1>
          <heading>Interfacing with the initscript system</heading>

          <p>
            Maintainers should use the abstraction layer provided by
            the <prgn>update-rc.d</prgn> and <prgn>invoke-rc.d</prgn>
            programs to deal with initscripts in their packages'
            scripts such as <prgn>postinst</prgn>, <prgn>prerm</prgn>
            and <prgn>postrm</prgn>.
          </p>

          <p>
            Directly managing the /etc/rc?.d links and directly
            invoking the <file>/etc/init.d/</file> initscripts should
            be done only by packages providing the initscript
            subsystem (such as <prgn>sysv-rc</prgn> and
            <prgn>file-rc</prgn>).
          </p>

          <sect2>
            <heading>Managing the links</heading>

            <p>
              The program <prgn>update-rc.d</prgn> is provided for
              package maintainers to arrange for the proper creation and
              removal of <file>/etc/rc<var>n</var>.d</file> symbolic links,
              or their functional equivalent if another method is being
              used.  This may be used by maintainers in their packages'
              <prgn>postinst</prgn> and <prgn>postrm</prgn> scripts.
            </p>

            <p>
              You must not include any <file>/etc/rc<var>n</var>.d</file>
              symbolic links in the actual archive or manually create or
              remove the symbolic links in maintainer scripts; you must
              use the <prgn>update-rc.d</prgn> program instead.  (The
              former will fail if an alternative method of maintaining
              runlevel information is being used.)  You must not include
              the <file>/etc/rc<var>n</var>.d</file> directories themselves
              in the archive either.  (Only the <tt>sysvinit</tt>
              package may do so.)
            </p>

            <p>
              By default <prgn>update-rc.d</prgn> will start services in
              each of the multi-user state runlevels (2, 3, 4, and 5)
              and stop them in the halt runlevel (0), the single-user
              runlevel (1) and the reboot runlevel (6).  The system
              administrator will have the opportunity to customize
              runlevels by simply adding, moving, or removing the
              symbolic links in <file>/etc/rc<var>n</var>.d</file> if
              symbolic links are being used, or by modifying
              <file>/etc/runlevel.conf</file> if the <tt>file-rc</tt> method
              is being used.
            </p>

            <p>
              To get the default behavior for your package, put in your
              <prgn>postinst</prgn> script
              <example compact="compact">
                update-rc.d <var>package</var> defaults
              </example>
              and in your <prgn>postrm</prgn>
              <example compact="compact">
                if [ "$1" = purge ]; then
                update-rc.d <var>package</var> remove
                fi
              </example>. Note that if your package changes runlevels
              or priority, you may have to remove and recreate the links,
              since otherwise the old links may persist. Refer to the
              documentation of <prgn>update-rc.d</prgn>.
            </p>

            <p>
              This will use a default sequence number of 20.  If it does
              not matter when or in which order the <file>init.d</file>
              script is run, use this default.  If it does, then you
              should talk to the maintainer of the <prgn>sysvinit</prgn>
              package or post to <tt>debian-devel</tt>, and they will
              help you choose a number.
            </p>

            <p>
              For more information about using <tt>update-rc.d</tt>,
              please consult its man page <manref name="update-rc.d"
                section="8">.
            </p>
          </sect2>

          <sect2>
            <heading>Running initscripts</heading>
            <p>
              The program <prgn>invoke-rc.d</prgn> is provided to make
              it easier for package maintainers to properly invoke an
              initscript, obeying runlevel and other locally-defined
              constraints that might limit a package's right to start,
              stop and otherwise manage services. This program may be
              used by maintainers in their packages' scripts.
            </p>

            <p>
              The package maintainer scripts must use
              <prgn>invoke-rc.d</prgn> to invoke the
              <file>/etc/init.d/*</file> initscripts, instead of
              calling them directly.
            </p>

            <p>
              By default, <prgn>invoke-rc.d</prgn> will pass any
              action requests (start, stop, reload, restart...) to the
              <file>/etc/init.d</file> script, filtering out requests
              to start or restart a service out of its intended
              runlevels.
            </p>

            <p>
              Most packages will simply need to change:
              <example compact="compact">/etc/init.d/&lt;package&gt;
              &lt;action&gt;</example> in their <prgn>postinst</prgn>
              and <prgn>prerm</prgn> scripts to:
              <example compact="compact">
        if which invoke-rc.d >/dev/null 2>&1; then
                invoke-rc.d <var>package</var> &lt;action&gt;
        else
                /etc/init.d/<var>package</var> &lt;action&gt;
        fi
              </example>
            </p>

            <p>
              A package should register its initscript services using
              <prgn>update-rc.d</prgn> before it tries to invoke them
              using <prgn>invoke-rc.d</prgn>. Invocation of
              unregistered services may fail.
            </p>

            <p>
              For more information about using
              <prgn>invoke-rc.d</prgn>, please consult its man page
              <manref name="invoke-rc.d" section="8">.
            </p>
          </sect2>
        </sect1>

        <sect1>
          <heading>Boot-time initialization</heading>

          <p>
            There used to be another directory, <file>/etc/rc.boot</file>,
            which contained scripts which were run once per machine
            boot. This has been deprecated in favour of links from
            <file>/etc/rcS.d</file> to files in <file>/etc/init.d</file> as
            described in <ref id="/etc/init.d">.  Packages must not
            place files in <file>/etc/rc.boot</file>.
          </p>
        </sect1>

        <sect1>
          <heading>Example</heading>

          <p>
            An example on which you can base your
            <file>/etc/init.d</file> scripts is found in
            <file>/etc/init.d/skeleton</file>.
          </p>

        </sect1>
      </sect>

      <sect>
        <heading>Console messages from <file>init.d</file> scripts</heading>

        <p>
          This section describes the formats to be used for messages
          written to standard output by the <file>/etc/init.d</file>
          scripts.  The intent is to improve the consistency of
          Debian's startup and shutdown look and feel.  For this
          reason, please look very carefully at the details.  We want
          the messages to have the same format in terms of wording,
          spaces, punctuation and case of letters.
        </p>

        <p>
          Here is a list of overall rules that should be used for
          messages generated by <file>/etc/init.d</file> scripts.  
        </p>

        <p>
          <list>
            <item>
                The message should fit in one line (fewer than 80
                characters), start with a capital letter and end with
                a period (<tt>.</tt>) and line feed (<tt>"\n"</tt>).
            </item>

            <item>
              If the script is performing some time consuming task in
              the background (not merely starting or stopping a
              program, for instance), an ellipsis (three dots:
              <tt>...</tt>) should be output to the screen, with no
              leading or tailing whitespace or line feeds.
            </item>

            <item>
              The messages should appear as if the computer is telling
              the user what it is doing (politely :-), but should not
                mention "it" directly.  For example, instead of:
                <example compact="compact">
I'm starting network daemons: nfsd mountd.
                </example>
                the message should say
                <example compact="compact">
Starting network daemons: nfsd mountd.
                </example>
            </item>
          </list>
        </p>

        <p>
          <tt>init.d</tt> script should use the following  standard
          message formats for the situations enumerated below.
        </p>

        <p>
          <list>
            <item>
              <p>When daemons are started</p>

              <p>
                If the script starts one or more daemons, the output
                should look like this (a single line, no leading
                spaces):
                <example compact="compact">
Starting <var>description</var>: <var>daemon-1</var> ... <var>daemon-n</var>.
                </example>
                The <var>description</var> should describe the
                subsystem the daemon or set of daemons are part of,
                while <var>daemon-1</var> up to <var>daemon-n</var>
                denote each daemon's name (typically the file name of
                the program).
              </p>

              <p>
                For example, the output of <file>/etc/init.d/lpd</file>
                would look like:
                <example compact="compact">
Starting printer spooler: lpd.
                </example>
              </p>

              <p>
                This can be achieved by saying
                <example compact="compact">
echo -n "Starting printer spooler: lpd"
start-stop-daemon --start --quiet --exec /usr/sbin/lpd
echo "."
                </example>
                in the script. If there are more than one daemon to
                start, the output should look like this:
                <example compact="compact">
echo -n "Starting remote file system services:"
echo -n " nfsd"; start-stop-daemon --start --quiet nfsd
echo -n " mountd"; start-stop-daemon --start --quiet mountd
echo -n " ugidd"; start-stop-daemon --start --quiet ugidd
echo "."
                </example>
                This makes it possible for the user to see what is
                happening and when the final daemon has been started.
                Care should be taken in the placement of white spaces:
                in the example above the system administrators can
                easily comment out a line if they don't want to start
                a specific daemon, while the displayed message still
                looks good.
              </p>
            </item>

            <item>
              <p>When a system parameter is being set</p>

              <p>
                If you have to set up different system parameters
                during the system boot, you should use this format:
                <example compact="compact">
Setting <var>parameter</var> to "<var>value</var>".
                </example>
              </p>

              <p>
                You can use a statement such as the following to get
                the quotes right:
                <example compact="compact">
echo "Setting DNS domainname to \"$domainname\"."
                </example>
              </p>

              <p>
                Note that the same symbol (<tt>"</tt>) <!-- " --> is used
                for the left and right quotation marks.  A grave accent
                (<tt>`</tt>) is not a quote character; neither is an
                apostrophe (<tt>'</tt>).
              </p>
            </item>

            <item>
              <p>When a daemon is stopped or restarted</p>

              <p>
                When you stop or restart a daemon, you should issue a
                message identical to the startup message, except that
                <tt>Starting</tt> is replaced with <tt>Stopping</tt>
                or <tt>Restarting</tt> respectively.
              </p>

              <p>
                For example, stopping the printer daemon will look like
                this:
                <example compact="compact">
Stopping printer spooler: lpd.
                </example>
              </p>
            </item>

            <item>
              <p>When something is executed</p>

              <p>
                There are several examples where you have to run a
                program at system startup or shutdown to perform a
                specific task, for example, setting the system's clock
                using <prgn>netdate</prgn> or killing all processes
                when the system shuts down.  Your message should look
                like this:
                <example compact="compact">
Doing something very useful...done.
                </example>
                You should print the <tt>done.</tt> immediately after
                the job has been completed, so that the user is
                informed why they have to wait.  You can get this
                behavior by saying
                <example compact="compact">
echo -n "Doing something very useful..."
do_something
echo "done."
                </example>
                in your script.
              </p>
            </item>

            <item>
              <p>When the configuration is reloaded</p>

              <p>
                When a daemon is forced to reload its configuration
                files you should use the following format:
                <example compact="compact">
Reloading <var>description</var> configuration...done.
                </example>
                where <var>description</var> is the same as in the
                daemon starting message.
              </p>
            </item>
          </list>
        </p>
      </sect>

      <sect id="cron-jobs">
        <heading>Cron jobs</heading>

        <p>
          Packages must not modify the configuration file
          <file>/etc/crontab</file>, and they must not modify the files in
          <file>/var/spool/cron/crontabs</file>.
        </p>

        <p>
          If a package wants to install a job that has to be executed via
          cron, it should place a file named as specified
          in <ref id="cron-files"> into one or more of the following
          directories:
          <example compact="compact">
/etc/cron.hourly
/etc/cron.daily
/etc/cron.weekly
/etc/cron.monthly
          </example>
          As these directory names imply, the files within them are
          executed on an hourly, daily, weekly, or monthly basis,
          respectively. The exact times are listed in
          <file>/etc/crontab</file>.
        </p>

        <p>
          All files installed in any of these directories must be
          scripts (e.g., shell scripts or Perl scripts) so that they
          can easily be modified by the local system administrator.
          In addition, they must be treated as configuration files.
        </p>

        <p>
          If a certain job has to be executed at some other frequency or
          at a specific time, the package should install a file in
          <file>/etc/cron.d</file> with a name as specified
          in <ref id="cron-files">.  This file uses the same syntax
          as <file>/etc/crontab</file> and is processed
          by <prgn>cron</prgn> automatically. The file must also be
          treated as a configuration file. (Note that entries in the
          <file>/etc/cron.d</file> directory are not handled by
          <prgn>anacron</prgn>. Thus, you should only use this
          directory for jobs which may be skipped if the system is not
          running.)
        </p>

        <p>
          Unlike <file>crontab</file> files described in the IEEE Std
          1003.1-2008 (POSIX.1) available from
          <url id="http://www.opengroup.org/onlinepubs/9699919799/"
               name="The Open Group">, the files in
          <file>/etc/cron.d</file> and the file
          <file>/etc/crontab</file> have seven fields; namely:
          <enumlist>
            <item>Minute [0,59]</item>
            <item>Hour [0,23]</item>
            <item>Day of the month [1,31]</item>
            <item>Month of the year [1,12]</item>
            <item>Day of the week ([0,6] with 0=Sunday)</item>
            <item>Username</item>
            <item>Command to be run</item>
          </enumlist>
          Ranges of numbers are allowed.  Ranges are two numbers
          separated with a hyphen.  The specified range is inclusive.
          Lists are allowed.  A list is a set of numbers (or ranges)
          separated by commas. Step values can be used in conjunction
          with ranges.
        </p>

        <p>
          The scripts or <tt>crontab</tt> entries in these directories should
          check if all necessary programs are installed before they
          try to execute them. Otherwise, problems will arise when a
          package was removed but not purged since configuration files
          are kept on the system in this situation.
        </p>

        <p>
          Any <tt>cron</tt> daemon must provide
          <file>/usr/bin/crontab</file> and support normal
          <tt>crontab</tt> entries as specified in POSIX. The daemon
          must also support names for days and months, ranges, and
          step values. It has to support <file>/etc/crontab</file>,
          and correctly execute the scripts in
          <file>/etc/cron.d</file>. The daemon must also correctly
          execute scripts in
          <file>/etc/cron.{hourly,daily,weekly,monthly}</file>.
        </p>

        <sect1 id="cron-files">
          <heading>Cron job file names</heading>

          <p>
            The file name of a cron job file should normally match the
            name of the package from which it comes.
          </p>

          <p>
            If a package supplies multiple cron job files files in the
            same directory, the file names should all start with the name
            of the package (possibly modified as described below) followed
            by a hyphen (<tt>-</tt>) and a suitable suffix.
          </p>

          <p>
            A cron job file name must not include any period or plus
            characters (<tt>.</tt> or <tt>+</tt>) characters as this will
            cause cron to ignore the file.  Underscores (<tt>_</tt>)
            should be used instead of <tt>.</tt> and <tt>+</tt>
            characters.
          </p>
        </sect1>
      </sect>

      <sect id="menus">
        <heading>Menus</heading>

        <p>
          The Debian <tt>menu</tt> package provides a standard
          interface between packages providing applications and
          <em>menu programs</em> (either X window managers or
          text-based menu programs such as <prgn>pdmenu</prgn>).
        </p>

        <p>
          All packages that provide applications that need not be
          passed any special command line arguments for normal
          operation should register a menu entry for those
          applications, so that users of the <tt>menu</tt> package
          will automatically get menu entries in their window
          managers, as well in shells like <tt>pdmenu</tt>.
        </p>

        <p>
          Menu entries should follow the current menu policy.
        </p>

        <p>
          The menu policy can be found in the <tt>menu-policy</tt>
          files in the <tt>debian-policy</tt> package.
          It is also available from the Debian web mirrors at
          <tt><url name="/doc/packaging-manuals/menu-policy/"
                id="http://www.debian.org/doc/packaging-manuals/menu-policy/"></tt>.
        </p>

        <p>
          Please also refer to the <em>Debian Menu System</em>
          documentation that comes with the <package>menu</package>
          package for information about how to register your
          applications.
        </p>
      </sect>

      <sect id="mime">
        <heading>Multimedia handlers</heading>

        <p>
          MIME (Multipurpose Internet Mail Extensions, RFCs 2045-2049)
          is a mechanism for encoding files and data streams and
          providing meta-information about them, in particular their
          type (e.g. audio or video) and format (e.g. PNG, HTML,
          MP3).
        </p>

        <p>
          Registration of MIME type handlers allows programs like mail
          user agents and web browsers to invoke these handlers to
          view, edit or display MIME types they don't support directly.
        </p>

        <p>
          Packages which provide programs to view/show/play, compose, edit or
          print MIME types should register them as such by placing a file in
          <manref name="mailcap" section="5"> format (RFC 1524) in the directory
          <file>/usr/lib/mime/packages/</file>.  The file name should be the
          binary package's name.
        </p>

        <p>
          The <package>mime-support</package> package provides the
          <prgn>update-mime</prgn> program, which integrates these
          registrations in the <file>/etc/mailcap</file> file, using dpkg
          triggers<footnote>
            Creating, modifying or removing a file in
            <file>/usr/lib/mime/packages/</file> using maintainer scripts will
            not activate the trigger.  In that case, it can be done by calling
            <tt>dpkg-trigger --no-await /usr/lib/mime/packages</tt> from
            the maintainer script after creating, modifying, or removing
            the file.
          </footnote>.
          Packages using this facility <em>should not</em> depend on,
          recommend, or suggest <prgn>mime-support</prgn>.
        </p>
      </sect>

      <sect>
        <heading>Keyboard configuration</heading>

        <p>
          To achieve a consistent keyboard configuration so that all
          applications interpret a keyboard event the same way, all
          programs in the Debian distribution must be configured to
          comply with the following guidelines.
        </p>

        <p>
          The following keys must have the specified interpretations:

          <taglist>
            <tag><tt>&lt;--</tt></tag>
            <item>delete the character to the left of the cursor</item>

            <tag><tt>Delete</tt></tag>
            <item>delete the character to the right of the cursor</item>

            <tag><tt>Control+H</tt></tag>
            <item>emacs: the help prefix</item>
          </taglist>

          The interpretation of any keyboard events should be
          independent of the terminal that is used, be it a virtual
          console, an X terminal emulator, an rlogin/telnet session,
          etc.
        </p>

        <p>
          The following list explains how the different programs
          should be set up to achieve this:
        </p>

        <p>
          <list>
            <item>
                <tt>&lt;--</tt> generates <tt>KB_BackSpace</tt> in X.
            </item>

            <item>
                <tt>Delete</tt> generates <tt>KB_Delete</tt> in X.
            </item>

            <item>
                X translations are set up to make
                <tt>KB_Backspace</tt> generate ASCII DEL, and to make
                <tt>KB_Delete</tt> generate <tt>ESC [ 3 ~</tt> (this
                is the vt220 escape code for the "delete character"
                key).  This must be done by loading the X resources
                using <prgn>xrdb</prgn> on all local X displays, not
                using the application defaults, so that the
                translation resources used correspond to the
                <prgn>xmodmap</prgn> settings.
            </item>

            <item>
                The Linux console is configured to make
                <tt>&lt;--</tt> generate DEL, and <tt>Delete</tt>
                generate <tt>ESC [ 3 ~</tt>.
            </item>

            <item>
                X applications are configured so that <tt>&lt;</tt>
                deletes left, and <tt>Delete</tt> deletes right.  Motif
                applications already work like this.
            </item>

            <item>
                Terminals should have <tt>stty erase ^?</tt> .
            </item>

            <item>
                The <tt>xterm</tt> terminfo entry should have <tt>ESC
                [ 3 ~</tt> for <tt>kdch1</tt>, just as for
                <tt>TERM=linux</tt> and <tt>TERM=vt220</tt>.
            </item>

            <item>
                Emacs is programmed to map <tt>KB_Backspace</tt> or
                the <tt>stty erase</tt> character to
                <tt>delete-backward-char</tt>, and <tt>KB_Delete</tt>
                or <tt>kdch1</tt> to <tt>delete-forward-char</tt>, and
                <tt>^H</tt> to <tt>help</tt> as always.
            </item>

            <item>
                Other applications use the <tt>stty erase</tt>
                character and <tt>kdch1</tt> for the two delete keys,
                with ASCII DEL being "delete previous character" and
                <tt>kdch1</tt> being "delete character under
                cursor".
            </item>

          </list>
        </p>

        <p>
          This will solve the problem except for the following
          cases:
        </p>

        <p>
          <list>
            <item>
                Some terminals have a <tt>&lt;--</tt> key that cannot
                be made to produce anything except <tt>^H</tt>.  On
                these terminals Emacs help will be unavailable on
                <tt>^H</tt> (assuming that the <tt>stty erase</tt>
                character takes precedence in Emacs, and has been set
                correctly).  <tt>M-x help</tt> or <tt>F1</tt> (if
                available) can be used instead.
            </item>

            <item>
                Some operating systems use <tt>^H</tt> for <tt>stty
                erase</tt>.  However, modern telnet versions and all
                rlogin versions propagate <tt>stty</tt> settings, and
                almost all UNIX versions honour <tt>stty erase</tt>.
                Where the <tt>stty</tt> settings are not propagated
                correctly, things can be made to work by using
                <tt>stty</tt> manually.
            </item>

            <item>
                Some systems (including previous Debian versions) use
                <prgn>xmodmap</prgn> to arrange for both
                <tt>&lt;--</tt> and <tt>Delete</tt> to generate
                <tt>KB_Delete</tt>.  We can change the behavior of
                their X clients using the same X resources that we use
                to do it for our own clients, or configure our clients
                using their resources when things are the other way
                around.  On displays configured like this
                <tt>Delete</tt> will not work, but <tt>&lt;--</tt>
                will.
            </item>

            <item>
                Some operating systems have different <tt>kdch1</tt>
                settings in their <tt>terminfo</tt> database for
                <tt>xterm</tt> and others.  On these systems the
                <tt>Delete</tt> key will not work correctly when you
                log in from a system conforming to our policy, but
                <tt>&lt;--</tt> will.
            </item>
          </list>
        </p>
      </sect>

      <sect>
        <heading>Environment variables</heading>

        <p>
          A program must not depend on environment variables to get
          reasonable defaults.  (That's because these environment
          variables would have to be set in a system-wide
          configuration file like <file>/etc/profile</file>, which is not
          supported by all shells.)
        </p>

        <p>
          If a program usually depends on environment variables for its
          configuration, the program should be changed to fall back to
          a reasonable default configuration if these environment
          variables are not present. If this cannot be done easily
          (e.g., if the source code of a non-free program is not
          available), the program must be replaced by a small
          "wrapper" shell script which sets the environment variables
          if they are not already defined, and calls the original program.
        </p>

        <p>
          Here is an example of a wrapper script for this purpose:

          <example compact="compact">
#!/bin/sh
BAR=${BAR:-/var/lib/fubar}
export BAR
exec /usr/lib/foo/foo "$@"
          </example>
        </p>

        <p>
          Furthermore, as <file>/etc/profile</file> is a configuration
          file of the <prgn>base-files</prgn> package, other packages must
          not put any environment variables or other commands into that
          file.
        </p>
      </sect>

      <sect id="doc-base">
        <heading>Registering Documents using doc-base</heading>

        <p>
          The <package>doc-base</package> package implements a
          flexible mechanism for handling and presenting
          documentation. The recommended practice is for every Debian
          package that provides online documentation (other than just
          manual pages) to register these documents with
          <package>doc-base</package> by installing a
          <package>doc-base</package> control file in
          <file>/usr/share/doc-base/</file>.
        </p> 
        <p>
          Please refer to the documentation that comes with the
          <package>doc-base</package>  package for information and
          details. 
        </p>
      </sect>

      <sect id="alternateinit">
        <heading>Alternate init systems</heading>
        <p>
          A number of other init systems are available now in Debian that
          can be used in place of <package>sysvinit</package>.  Alternative
          init implementations must support running SysV init scripts as
          described at <ref id="sysvinit"> for compatibility.
        </p>
        <p>
          Packages may integrate with these replacement init systems by
          providing implementation-specific configuration information about
          how and when to start a service or in what order to run certain
          tasks at boot time.  However, any package integrating with other
          init systems must also be backwards-compatible with
          <package>sysvinit</package> by providing a SysV-style init script
          with the same name as and equivalent functionality to any
          init-specific job, as this is the only start-up configuration
          method guaranteed to be supported by all init implementations.  An
          exception to this rule is scripts or jobs provided by the init
          implementation itself; such jobs may be required for an
          implementation-specific equivalent of the <file>/etc/rcS.d/</file>
          scripts and may not have a one-to-one correspondence with the init
          scripts.
        </p>
        <sect1 id="upstart">
          <heading>Event-based boot with upstart</heading>

          <p>
            Packages may integrate with the <prgn>upstart</prgn> event-based
            boot system by installing job files in the
            <file>/etc/init</file> directory.  SysV init scripts for which
            an equivalent upstart job is available must query the output of
            the command <prgn>initctl version</prgn> for the string
            <tt>upstart</tt> and avoid running in favor of the native
            upstart job, using a test such as this:
            <example compact="compact">
if [ "$1" = start ] && which initctl >/dev/null && initctl version | grep -q upstart
then
        exit 1
fi
            </example>
          </p>
          <p>
            Because packages shipping upstart jobs may be installed on
            systems that are not using upstart, maintainer scripts must
            still use the common <prgn>update-rc.d</prgn> and
            <prgn>invoke-rc.d</prgn> interfaces for configuring runlevels
            and for starting and stopping services.  These maintainer
            scripts must not call the upstart <prgn>start</prgn>,
            <prgn>restart</prgn>, <prgn>reload</prgn>, or <prgn>stop</prgn>
            interfaces directly.  Instead, implementations of
            <prgn>invoke-rc.d</prgn> must detect when upstart is running and
            when an upstart job with the same name as an init script is
            present, and perform the requested action using the upstart job
            instead of the init script.
          </p>
          <p>
            Dependency-based boot managers for SysV init scripts, such as
            <prgn>startpar</prgn>, may avoid running a given init script
            entirely when an equivalent upstart job is present, to avoid
            unnecessary forking of no-op init scripts.  In this case, the
            boot manager should integrate with upstart to detect when the
            upstart job in question is started or stopped to know when the
            dependency has been satisfied.
          </p>
         </sect1>
      </sect>

    </chapt>


    <chapt id="files">
      <heading>Files</heading>

      <sect id="binaries">
        <heading>Binaries</heading>

        <p>
          Two different packages must not install programs with
          different functionality but with the same filenames.  (The
          case of two programs having the same functionality but
          different implementations is handled via "alternatives" or
          the "Conflicts" mechanism.  See <ref id="maintscripts"> and
          <ref id="conflicts"> respectively.)  If this case happens,
          one of the programs must be renamed.  The maintainers should
          report this to the <tt>debian-devel</tt> mailing list and
          try to find a consensus about which program will have to be
          renamed.  If a consensus cannot be reached, <em>both</em>
          programs must be renamed.
        </p>
        <p>
          Binary executables must not be statically linked with the GNU C
          library, since this prevents the binary from benefiting from
          fixes and improvements to the C library without being rebuilt
          and complicates security updates.  This requirement may be
          relaxed for binary executables whose intended purpose is to
          diagnose and fix the system in situations where the GNU C
          library may not be usable (such as system recovery shells or
          utilities like ldconfig) or for binary executables where the
          security benefits of static linking outweigh the drawbacks.
        </p>
        <p>
         By default, when a package is being built, any binaries
         created should include debugging information, as well as
         being compiled with optimization.  You should also turn on
         as many reasonable compilation warnings as possible; this
         makes life easier for porters, who can then look at build
         logs for possible problems.  For the C programming language,
         this means the following compilation parameters should be
         used:
          <example compact="compact">
CC = gcc
CFLAGS = -O2 -g -Wall # sane warning options vary between programs
LDFLAGS = # none
INSTALL = install -s # (or use strip on the files in debian/tmp)
          </example>
        </p>

        <p>
          Note that by default all installed binaries should be stripped,
          either by using the <tt>-s</tt> flag to
          <prgn>install</prgn>, or by calling <prgn>strip</prgn> on
          the binaries after they have been copied into
          <file>debian/tmp</file> but before the tree is made into a
          package.
        </p>

        <p>
          Although binaries in the build tree should be compiled with
          debugging information by default, it can often be difficult to
          debug programs if they are also subjected to compiler
          optimization.  For this reason, it is recommended to support the
          standardized environment variable <tt>DEB_BUILD_OPTIONS</tt>
          (see <ref id="debianrules-options">).  This variable can contain
          several flags to change how a package is compiled and built.
        </p>

        <p>
          It is up to the package maintainer to decide what
          compilation options are best for the package.  Certain
          binaries (such as computationally-intensive programs) will
          function better with certain flags (<tt>-O3</tt>, for
          example); feel free to use them.  Please use good judgment
          here.  Don't use flags for the sake of it; only use them
          if there is good reason to do so.  Feel free to override
          the upstream author's ideas about which compilation
          options are best: they are often inappropriate for our
          environment.
        </p>
      </sect>


      <sect id="libraries">
        <heading>Libraries</heading>

        <p>
          If the package is <strong>architecture: any</strong>, then
          the shared library compilation and linking flags must have
          <tt>-fPIC</tt>, or the package shall not build on some of
          the supported architectures<footnote>
            <p>
              If you are using GCC, <tt>-fPIC</tt> produces code with
              relocatable position independent code, which is required for
              most architectures to create a shared library, with i386 and
              perhaps some others where non position independent code is
              permitted in a shared library.
            </p>
            <p>
              Position independent code may have a performance penalty,
              especially on <tt>i386</tt>. However, in most cases the
              speed penalty must be measured against the memory wasted on
              the few architectures where non position independent code is
              even possible.
            </p>
          </footnote>. Any exception to this rule must be discussed on
          the mailing list <em>debian-devel@lists.debian.org</em>, and
          a rough consensus obtained.  The reasons for not compiling
          with <tt>-fPIC</tt> flag must be recorded in the file
          <tt>README.Debian</tt>, and care must be taken to either
          restrict the architecture or arrange for <tt>-fPIC</tt> to
          be used on architectures where it is required.<footnote>
            <p>
              Some of the reasons why this might be required is if the
              library contains hand crafted assembly code that is not
              relocatable, the speed penalty is excessive for compute
              intensive libs, and similar reasons.
            </p>
          </footnote>
        </p>
        <p>
          As to the static libraries, the common case is not to have
          relocatable code, since there is no benefit, unless in specific
          cases; therefore the static version must not be compiled
          with the <tt>-fPIC</tt> flag. Any exception to this rule
          should be discussed on the mailing list
          <em>debian-devel@lists.debian.org</em>,  and the reasons for
          compiling with the <tt>-fPIC</tt> flag must be recorded in
          the file <tt>README.Debian</tt>. <footnote>
            <p>
              Some of the reasons for linking static libraries with
              the <tt>-fPIC</tt> flag are if, for example, one needs a
              Perl API for a library that is under rapid development,
              and has an unstable API, so shared libraries are
              pointless at this phase of the library's development. In
              that case, since Perl needs a library with relocatable
              code, it may make sense to create a static library with
              relocatable code. Another reason cited is if you are
              distilling various libraries into a common shared
              library, like <tt>mklibs</tt> does in the Debian
              installer project.
            </p>
          </footnote>
        </p>
        <p>
          In other words, if both a shared and a static library is
          being built, each source unit (<tt>*.c</tt>, for example,
          for C files) will need to be compiled twice, for the normal
          case. 
        </p>

        <p>
          Libraries should be built with threading support and to be
          thread-safe if the library supports this.
        </p>

        <p>
          Although not enforced by the build tools, shared libraries
          must be linked against all libraries that they use symbols from
          in the same way that binaries are.  This ensures the correct
          functioning of the <qref id="sharedlibs-symbols">symbols</qref>
          and <qref id="sharedlibs-shlibdeps">shlibs</qref>
          systems and guarantees that all libraries can be safely opened
          with <tt>dlopen()</tt>.  Packagers may wish to use the gcc
          option <tt>-Wl,-z,defs</tt> when building a shared library.
          Since this option enforces symbol resolution at build time,
          a missing library reference will be caught early as a fatal
          build error.
        </p>

        <p>
          All installed shared libraries should be stripped with
          <example compact="compact">
strip --strip-unneeded <var>your-lib</var>
          </example>
          (The option <tt>--strip-unneeded</tt> makes
          <prgn>strip</prgn> remove only the symbols which aren't
          needed for relocation processing.)  Shared libraries can
          function perfectly well when stripped, since the symbols for
          dynamic linking are in a separate part of the ELF object
          file.<footnote>
              You might also want to use the options
              <tt>--remove-section=.comment</tt> and
              <tt>--remove-section=.note</tt> on both shared libraries
              and executables, and <tt>--strip-debug</tt> on static
              libraries.
          </footnote>
        </p>

        <p>
          Note that under some circumstances it may be useful to
          install a shared library unstripped, for example when
          building a separate package to support debugging.
        </p>

        <p>
          Shared object files (often <file>.so</file> files) that are not
          public libraries, that is, they are not meant to be linked
          to by third party executables (binaries of other packages),
          should be installed in subdirectories of the
          <file>/usr/lib</file> directory.  Such files are exempt from the
          rules that govern ordinary shared libraries, except that
          they must not be installed executable and should be
          stripped.<footnote>
              A common example are the so-called "plug-ins",
              internal shared objects that are dynamically loaded by
              programs using <manref name="dlopen" section="3">.
          </footnote>
        </p>

        <p>
          Packages that use <prgn>libtool</prgn> to create and install
          their shared libraries install a file containing additional
          metadata (ending in <file>.la</file>) alongside the library.
          For public libraries intended for use by other packages, these
          files normally should not be included in the Debian package,
          since the information they include is not necessary to link with
          the shared library on Debian and can add unnecessary additional
          dependencies to other programs or libraries.<footnote>
            These files store, among other things, all libraries on which
            that shared library depends.  Unfortunately, if
            the <file>.la</file> file is present and contains that
            dependency information, using <prgn>libtool</prgn> when
            linking against that library will cause the resulting program
            or library to be linked against those dependencies as well,
            even if this is unnecessary.  This can create unneeded
            dependencies on shared library packages that would otherwise
            be hidden behind the library ABI, and can make library
            transitions to new SONAMEs unnecessarily complicated and
            difficult to manage.
          </footnote>
          If the <file>.la</file> file is required for that library (if,
          for instance, it's loaded via <tt>libltdl</tt> in a way that
          requires that meta-information), the <tt>dependency_libs</tt>
          setting in the <file>.la</file> file should normally be set to
          the empty string.  If the shared library development package has
          historically included the <file>.la</file>, it must be retained
          in the development package (with <tt>dependency_libs</tt>
          emptied) until all libraries that depend on it have removed or
          emptied <tt>dependency_libs</tt> in their <file>.la</file>
          files to prevent linking with those other libraries
          using <prgn>libtool</prgn> from failing.
        </p>

        <p>
          If the <file>.la</file> must be included, it should be included
          in the development (<tt>-dev</tt>) package, unless the library
          will be loaded by <prgn>libtool</prgn>'s <tt>libltdl</tt>
          library.  If it is intended for use with <tt>libltdl</tt>,
          the <file>.la</file> files must go in the run-time library
          package.
        </p>

        <p>
          These requirements for handling of <file>.la</file> files do not
          apply to loadable modules or libraries not installed in
          directories searched by default by the dynamic linker.  Packages
          installing loadable modules will frequently need to install
          the <file>.la</file> files alongside the modules so that they
          can be loaded by <tt>libltdl</tt>.  <tt>dependency_libs</tt>
          does not need to be modified for libraries or modules that are
          not installed in directories searched by the dynamic linker by
          default and not intended for use by other packages.
        </p>

        <p>
          You must make sure that you use only released versions of
          shared libraries to build your packages; otherwise other
          users will not be able to run your binaries
          properly. Producing source packages that depend on
          unreleased compilers is also usually a bad
          idea.
        </p>
      </sect>


      <sect>
        <heading>Shared libraries</heading>
        <p>
          This section has moved to <ref id="sharedlibs">.
        </p>
      </sect>


      <sect id="scripts">
        <heading>Scripts</heading>

        <p>
          All command scripts, including the package maintainer
          scripts inside the package and used by <prgn>dpkg</prgn>,
          should have a <tt>#!</tt> line naming the shell to be used
          to interpret them.
        </p>

        <p>
          In the case of Perl scripts this should be
          <tt>#!/usr/bin/perl</tt>.
        </p>

        <p>
          When scripts are installed into a directory in the system
          PATH, the script name should not include an extension such
          as <tt>.sh</tt> or <tt>.pl</tt> that denotes the scripting
          language currently used to implement it.
        </p>
        <p>
          Shell scripts (<prgn>sh</prgn> and <prgn>bash</prgn>) other than
          <file>init.d</file> scripts should almost certainly start
          with <tt>set -e</tt> so that errors are detected.
          <file>init.d</file> scripts are something of a special case, due
          to how frequently they need to call commands that are allowed to
          fail, and it may instead be easier to check the exit status of
          commands directly.  See <ref id="writing-init"> for more
          information about writing <file>init.d</file> scripts.
        </p>
        <p>
          Every script should use <tt>set -e</tt> or check the exit status
          of <em>every</em> command.
        </p>
        <p>
          Scripts may assume that <file>/bin/sh</file> implements the
          SUSv3 Shell Command Language<footnote>
            Single UNIX Specification, version 3, which is also IEEE
            1003.1-2004 (POSIX), and is available on the World Wide Web
            from <url id="http://www.unix.org/version3/online.html"
                      name="The Open Group"> after free
            registration.</footnote>
          plus the following additional features not mandated by
          SUSv3:<footnote>
            These features are in widespread use in the Linux community
            and are implemented in all of bash, dash, and ksh, the most
            common shells users may wish to use as <file>/bin/sh</file>.
          </footnote>
          <list>
            <item><tt>echo -n</tt>, if implemented as a shell built-in,
              must not generate a newline.</item>
            <item><tt>test</tt>, if implemented as a shell built-in, must
              support <tt>-a</tt> and <tt>-o</tt> as binary logical
              operators.</item>
            <item><tt>local</tt> to create a scoped variable must be
              supported, including listing multiple variables in a single
              local command and assigning a value to a variable at the
              same time as localizing it.  <tt>local</tt> may or
              may not preserve the variable value from an outer scope if
              no assignment is present.  Uses such as:
<example compact>
fname () {
    local a b c=delta d
    # ... use a, b, c, d ...
}
</example>
              must be supported and must set the value of <tt>c</tt> to
              <tt>delta</tt>.
            </item>
            <item>The XSI extension to <prgn>kill</prgn> allowing <tt>kill
              -<var>signal</var></tt>, where <var>signal</var> is either
              the name of a signal or one of the numeric signals listed in
              the XSI extension (0, 1, 2, 3, 6, 9, 14, and 15), must be
              supported if <prgn>kill</prgn> is implemented as a shell
              built-in.
            </item>
            <item>The XSI extension to <prgn>trap</prgn> allowing numeric
              signals must be supported.  In addition to the signal
              numbers listed in the extension, which are the same as for
              <prgn>kill</prgn> above, 13 (SIGPIPE) must be allowed.
            </item>
          </list>
          If a shell script requires non-SUSv3 features from the shell
          interpreter other than those listed above, the appropriate shell
          must be specified in the first line of the script (e.g.,
          <tt>#!/bin/bash</tt>) and the package must depend on the package
          providing the shell (unless the shell package is marked
          "Essential", as in the case of <prgn>bash</prgn>).
        </p>

        <p>
          You may wish to restrict your script to SUSv3 features plus the
          above set when possible so that it may use <file>/bin/sh</file>
          as its interpreter.  Checking your script
          with <prgn>checkbashisms</prgn> from
          the <package>devscripts</package> package or running your script
          with an alternate shell such as <prgn>posh</prgn> may help
          uncover violations of the above requirements.  If in doubt
          whether a script complies with these requirements,
          use <file>/bin/bash</file>.
        </p>

        <p>
          Perl scripts should check for errors when making any
          system calls, including <tt>open</tt>, <tt>print</tt>,
          <tt>close</tt>, <tt>rename</tt> and <tt>system</tt>.
        </p>

        <p>
          <prgn>csh</prgn> and <prgn>tcsh</prgn> should be avoided as
          scripting languages.  See <em>Csh Programming Considered
          Harmful</em>, one of the <tt>comp.unix.*</tt> FAQs, which
          can be found at <url id="http://www.faqs.org/faqs/unix-faq/shell/csh-whynot/">.
          If an upstream package comes with <prgn>csh</prgn> scripts
          then you must make sure that they start with
          <tt>#!/bin/csh</tt> and make your package depend on the
          <prgn>c-shell</prgn> virtual package.
        </p>

        <p>
          Any scripts which create files in world-writeable
          directories (e.g., in <file>/tmp</file>) must use a
          mechanism which will fail atomically if a file with the same
          name already exists.
        </p>

        <p>
          The Debian base system provides the <prgn>tempfile</prgn>
          and <prgn>mktemp</prgn> utilities for use by scripts for
          this purpose.
        </p>
      </sect>


      <sect>
        <heading>Symbolic links</heading>

        <p>
          In general, symbolic links within a top-level directory should
          be relative, and symbolic links pointing from one top-level
          directory to or into another should be absolute. (A top-level
          directory is a sub-directory of the root
          directory <file>/</file>.)  For example, a symbolic link
          from <file>/usr/lib/foo</file> to <file>/usr/share/bar</file>
          should be relative (<file>../share/bar</file>), but a symbolic
          link from <file>/var/run</file> to <file>/run</file> should be
          absolute.<footnote>
            This is necessary to allow top-level directories to be
            symlinks.  If linking <file>/var/run</file>
            to <file>/run</file> were done with the relative symbolic
            link <file>../run</file>, but <file>/var</file> were a
            symbolic link to <file>/srv/disk1</file>, the symbolic link
            would point to <file>/srv/run</file> rather than the intended
            target.
          </footnote>
        </p>

        <p>
          In addition, symbolic links should be specified as short as
          possible, i.e., link targets like <file>foo/../bar</file> are
          deprecated.
        </p>

        <p>
          Note that when creating a relative link using
          <prgn>ln</prgn> it is not necessary for the target of the
          link to exist relative to the working directory you're
          running <prgn>ln</prgn> from, nor is it necessary to change
          directory to the directory where the link is to be made.
          Simply include the string that should appear as the target
          of the link (this will be a pathname relative to the
          directory in which the link resides) as the first argument
          to <prgn>ln</prgn>.
        </p>

        <p>
          For example, in your <prgn>Makefile</prgn> or
          <file>debian/rules</file>, you can do things like:
          <example compact="compact">
ln -fs gcc $(prefix)/bin/cc
ln -fs gcc debian/tmp/usr/bin/cc
ln -fs ../sbin/sendmail $(prefix)/bin/runq
ln -fs ../sbin/sendmail debian/tmp/usr/bin/runq
          </example>
        </p>

        <p>
          A symbolic link pointing to a compressed file (in the sense
          that it is meant to be uncompressed with <prgn>unzip</prgn>
          or <prgn>zless</prgn> etc.) should always
          have the same file extension as the referenced file. (For
          example, if a file <file>foo.gz</file> is referenced by a
          symbolic link, the filename of the link has to end with
          "<file>.gz</file>" too, as in <file>bar.gz</file>.)
        </p>
      </sect>

      <sect>
        <heading>Device files</heading>

        <p>
          Packages must not include device files or named pipes in the
          package file tree.
        </p>

        <p>
          If a package needs any special device files that are not
          included in the base system, it must call
          <prgn>MAKEDEV</prgn> in the <prgn>postinst</prgn> script,
          after notifying the user<footnote>
              This notification could be done via a (low-priority)
              debconf message, or an echo (printf) statement.
          </footnote>.
        </p>

        <p>
          Packages must not remove any device files in the
          <prgn>postrm</prgn> or any other script. This is left to the
          system administrator.
        </p>

        <p>
          Debian uses the serial devices
          <file>/dev/ttyS*</file>. Programs using the old
          <file>/dev/cu*</file> devices should be changed to use
          <file>/dev/ttyS*</file>.
        </p>

        <p>
          Named pipes needed by the package must be created in
          the <prgn>postinst</prgn> script<footnote>
            It's better to use <prgn>mkfifo</prgn> rather
            than <prgn>mknod</prgn> to create named pipes so that
            automated checks for packages incorrectly creating device
            files with <prgn>mknod</prgn> won't have false positives.
          </footnote> and removed in
          the <prgn>prerm</prgn> or <prgn>postrm</prgn> script as
          appropriate.
        </p>
      </sect>

      <sect id="config-files">
        <heading>Configuration files</heading>

        <sect1>
          <heading>Definitions</heading>

          <p>
            <taglist>
              <tag>configuration file</tag>
              <item>
                  A file that affects the operation of a program, or
                  provides site- or host-specific information, or
                  otherwise customizes the behavior of a program.
                  Typically, configuration files are intended to be
                  modified by the system administrator (if needed or
                  desired) to conform to local policy or to provide
                  more useful site-specific behavior.
              </item>

              <tag><tt>conffile</tt></tag>
              <item>
                  A file listed in a package's <tt>conffiles</tt>
                  file, and is treated specially by <prgn>dpkg</prgn>
                  (see <ref id="configdetails">).
              </item>
            </taglist>
          </p>

          <p>
            The distinction between these two is important; they are
            not interchangeable concepts. Almost all
            <tt>conffile</tt>s are configuration files, but many
            configuration files are not <tt>conffiles</tt>.
          </p>

          <p>
            As noted elsewhere, <file>/etc/init.d</file> scripts,
            <file>/etc/default</file> files, scripts installed in
            <file>/etc/cron.{hourly,daily,weekly,monthly}</file>, and cron
            configuration installed in <file>/etc/cron.d</file> must be
            treated as configuration files.  In general, any script that
            embeds configuration information is de-facto a configuration
            file and should be treated as such.
          </p>
        </sect1>

        <sect1>
          <heading>Location</heading>

          <p>
            Any configuration files created or used by your package
            must reside in <file>/etc</file>. If there are several,
            consider creating a subdirectory of <file>/etc</file>
            named after your package.
          </p>

          <p>
            If your package creates or uses configuration files
            outside of <file>/etc</file>, and it is not feasible to modify
            the package to use <file>/etc</file> directly, put the files
            in <file>/etc</file> and create symbolic links to those files
            from the location that the package requires.
          </p>
        </sect1>

        <sect1>
          <heading>Behavior</heading>

          <p>
            Configuration file handling must conform to the following
            behavior:
            <list compact="compact">
              <item>
                  local changes must be preserved during a package
                  upgrade, and
              </item>
              <item>
                  configuration files must be preserved when the
                  package is removed, and only deleted when the
                  package is purged.
              </item>
            </list>
            Obsolete configuration files without local changes should be
            removed by the package during upgrade.<footnote>
              The <prgn>dpkg-maintscript-helper</prgn> tool, available from the
              <package>dpkg</package> package, can help for this task.</footnote>
          </p>

          <p>
            The easy way to achieve this behavior is to make the
            configuration file a <tt>conffile</tt>. This is
            appropriate only if it is possible to distribute a default
            version that will work for most installations, although
            some system administrators may choose to modify it. This
            implies that the default version will be part of the
            package distribution, and must not be modified by the
            maintainer scripts during installation (or at any other
            time).
          </p>

          <p>
            In order to ensure that local changes are preserved
            correctly, no package may contain or make hard links to
            conffiles.<footnote>
                Rationale: There are two problems with hard links.
                The first is that some editors break the link while
                editing one of the files, so that the two files may
                unwittingly become unlinked and different.  The second
                is that <prgn>dpkg</prgn> might break the hard link
                while upgrading <tt>conffile</tt>s.
            </footnote>
          </p>

          <p>
            The other way to do it is via the maintainer scripts.  In
            this case, the configuration file must not be listed as a
            <tt>conffile</tt> and must not be part of the package
            distribution. If the existence of a file is required for
            the package to be sensibly configured it is the
            responsibility of the package maintainer to provide
            maintainer scripts which correctly create, update and
            maintain the file and remove it on purge.  (See <ref
            id="maintainerscripts"> for more information.)  These
            scripts must be idempotent (i.e., must work correctly if
            <prgn>dpkg</prgn> needs to re-run them due to errors
            during installation or removal), must cope with all the
            variety of ways <prgn>dpkg</prgn> can call maintainer
            scripts, must not overwrite or otherwise mangle the user's
            configuration without asking, must not ask unnecessary
            questions (particularly during upgrades), and must
            otherwise be good citizens.
          </p>

          <p>
            The scripts are not required to configure every possible
            option for the package, but only those necessary to get
            the package running on a given system. Ideally the
            sysadmin should not have to do any configuration other
            than that done (semi-)automatically by the
            <prgn>postinst</prgn> script.
          </p>

          <p>
            A common practice is to create a script called
            <file><var>package</var>-configure</file> and have the
            package's <prgn>postinst</prgn> call it if and only if the
            configuration file does not already exist.  In certain
            cases it is useful for there to be an example or template
            file which the maintainer scripts use.  Such files should
            be in <file>/usr/share/<var>package</var></file> or
            <file>/usr/lib/<var>package</var></file> (depending on whether
            they are architecture-independent or not).  There should
            be symbolic links to them from
            <file>/usr/share/doc/<var>package</var>/examples</file> if
            they are examples, and should be perfectly ordinary
            <prgn>dpkg</prgn>-handled files (<em>not</em>
            configuration files).
          </p>

          <p>
            These two styles of configuration file handling must
            not be mixed, for that way lies madness:
            <prgn>dpkg</prgn> will ask about overwriting the file
            every time the package is upgraded.
          </p>
        </sect1>

        <sect1>
          <heading>Sharing configuration files</heading>

          <p>
            If two or more packages use the same configuration file
            and it is reasonable for both to be installed at the same
            time, one of these packages must be defined as
            <em>owner</em> of the configuration file, i.e., it will be
            the package which handles that file as a configuration
            file.  Other packages that use the configuration file must
            depend on the owning package if they require the
            configuration file to operate. If the other package will
            use the configuration file if present, but is capable of
            operating without it, no dependency need be declared.
          </p>

          <p>
            If it is desirable for two or more related packages to
            share a configuration file <em>and</em> for all of the
            related packages to be able to modify that configuration
            file, then the following should be done:
            <enumlist compact="compact">
              <item>
                  One of the related packages (the "owning" package)
                  will manage the configuration file with maintainer
                  scripts as described in the previous section.
              </item>
              <item>
                  The owning package should also provide a program
                  that the other packages may use to modify the
                  configuration file.
              </item>
              <item>
                  The related packages must use the provided program
                  to make any desired modifications to the
                  configuration file.  They should either depend on
                  the core package to guarantee that the configuration
                  modifier program is available or accept gracefully
                  that they cannot modify the configuration file if it
                  is not.  (This is in addition to the fact that the
                  configuration file may not even be present in the
                  latter scenario.)
              </item>
            </enumlist>
          </p>

          <p>
            Sometimes it's appropriate to create a new package which
            provides the basic infrastructure for the other packages
            and which manages the shared configuration files.  (The
            <tt>sgml-base</tt> package is a good example.)
          </p>

          <p>
            If the configuration file cannot be shared as described above,
            the packages must be marked as conflicting with each other.
            Two packages that specify the same file as
            a <tt>conffile</tt> must conflict.  This is an instance of the
            general rule about not sharing files.  Neither alternatives
            nor diversions are likely to be appropriate in this case; in
            particular, <prgn>dpkg</prgn> does not handle diverted
            <tt>conffile</tt>s well.
          </p>

          <p>
            When two packages both declare the same <tt>conffile</tt>, they
            may see left-over configuration files from each other even
            though they conflict with each other.  If a user removes
            (without purging) one of the packages and installs the other,
            the new package will take over the <tt>conffile</tt> from the
            old package.  If the file was modified by the user, it will be
            treated the same as any other locally
            modified <tt>conffile</tt> during an upgrade.
          </p>

          <p>
            The maintainer scripts must not alter a <tt>conffile</tt>
            of <em>any</em> package, including the one the scripts
            belong to.
          </p>
        </sect1>

        <sect1>
          <heading>User configuration files ("dotfiles")</heading>

          <p>
            The files in <file>/etc/skel</file> will automatically be
            copied into new user accounts by <prgn>adduser</prgn>.
            No other program should reference the files in
            <file>/etc/skel</file>.
          </p>

          <p>
            Therefore, if a program needs a dotfile to exist in
            advance in <file>$HOME</file> to work sensibly, that dotfile
            should be installed in <file>/etc/skel</file> and treated as a
            configuration file.
          </p>

          <p>
            However, programs that require dotfiles in order to
            operate sensibly are a bad thing, unless they do create
            the dotfiles themselves automatically.
          </p>

          <p>
            Furthermore, programs should be configured by the Debian
            default installation to behave as closely to the upstream
            default behavior as possible.
          </p>

          <p>
            Therefore, if a program in a Debian package needs to be
            configured in some way in order to operate sensibly, that
            should be done using a site-wide configuration file placed
            in <file>/etc</file>.  Only if the program doesn't support a
            site-wide default configuration and the package maintainer
            doesn't have time to add it may a default per-user file be
            placed in <file>/etc/skel</file>.
          </p>

          <p>
            <file>/etc/skel</file> should be as empty as we can make it.
            This is particularly true because there is no easy (or
            necessarily desirable) mechanism for ensuring that the
            appropriate dotfiles are copied into the accounts of
            existing users when a package is installed.
          </p>
        </sect1>
      </sect>

      <sect>
        <heading>Log files</heading>
        <p>
          Log files should usually be named
          <file>/var/log/<var>package</var>.log</file>.  If you have many
          log files, or need a separate directory for permission
          reasons (<file>/var/log</file> is writable only by
          <file>root</file>), you should usually create a directory named
          <file>/var/log/<var>package</var></file> and place your log
          files there.
        </p>

        <p>
          Log files must be rotated occasionally so that they don't grow
          indefinitely.  The best way to do this is to install a log
          rotation configuration file in the
          directory <file>/etc/logrotate.d</file>, normally
          named <file>/etc/logrotate.d/<var>package</var></file>, and use
          the facilities provided by <prgn>logrotate</prgn>.
          <footnote>
            <p>
              The traditional approach to log files has been to set up
              <em>ad hoc</em> log rotation schemes using simple shell
              scripts and cron.  While this approach is highly
              customizable, it requires quite a lot of sysadmin work.
              Even though the original Debian system helped a little
              by automatically installing a system which can be used
              as a template, this was deemed not enough.
            </p>

            <p>
              The use of <prgn>logrotate</prgn>, a program developed
              by Red Hat, is better, as it centralizes log management.
              It has both a configuration file
              (<file>/etc/logrotate.conf</file>) and a directory where
              packages can drop their individual log rotation
              configurations (<file>/etc/logrotate.d</file>).
            </p>
          </footnote>
          Here is a good example for a logrotate config
          file (for more information see <manref name="logrotate"
            section="8">):
          <example compact="compact">
/var/log/foo/*.log {
    rotate 12
    weekly
    compress
    missingok
    postrotate
        start-stop-daemon -K -p /var/run/foo.pid -s HUP -x /usr/sbin/foo -q
    endscript
}
          </example>
          This rotates all files under <file>/var/log/foo</file>, saves 12
          compressed generations, and tells the daemon to reopen its log
          files after the log rotation.  It skips this log rotation
          (via <tt>missingok</tt>) if no such log file is present, which
          avoids errors if the package is removed but not purged.
        </p>

        <p>
          Log files should be removed when the package is
          purged (but not when it is only removed).  This should be
          done by the <prgn>postrm</prgn> script when it is called
          with the argument <tt>purge</tt> (see <ref
          id="removedetails">).
        </p>
      </sect>

      <sect id="permissions-owners">
        <heading>Permissions and owners</heading>

        <p>
          The rules in this section are guidelines for general use.
          If necessary you may deviate from the details below.
          However, if you do so you must make sure that what is done
          is secure and you should try to be as consistent as possible
          with the rest of the system.  You should probably also
          discuss it on <prgn>debian-devel</prgn> first.
        </p>

        <p>
          Files should be owned by <tt>root:root</tt>, and made
          writable only by the owner and universally readable (and
          executable, if appropriate), that is mode 644 or 755.
        </p>

        <p>
          Directories should be mode 755 or (for group-writability)
          mode 2775.  The ownership of the directory should be
          consistent with its mode: if a directory is mode 2775, it
          should be owned by the group that needs write access to
          it.<footnote>
            <p>
              When a package is upgraded, and the owner or permissions
              of a file included in the package has changed, dpkg
              arranges for the ownership and permissions to be
              correctly set upon installation. However, this does not
              extend to directories; the permissions and ownership of
              directories already on the system does not change on
              install or upgrade of packages.  This makes sense, since
              otherwise common directories like <tt>/usr</tt> would
              always be in flux.  To correctly change permissions of a
              directory the package owns, explicit action is required,
              usually in the <tt>postinst</tt> script. Care must be
              taken to handle downgrades as well, in that case.
            </p>
          </footnote>
        </p>

        <p>
          Control information files should be owned by <tt>root:root</tt>
          and either mode 644 (for most files) or mode 755 (for
          executables such as <qref id="maintscripts">maintainer
          scripts</qref>).
        </p>

        <p>
          Setuid and setgid executables should be mode 4755 or 2755
          respectively, and owned by the appropriate user or group.
          They should not be made unreadable (modes like 4711 or
          2711 or even 4111); doing so achieves no extra security,
          because anyone can find the binary in the freely available
          Debian package; it is merely inconvenient.  For the same
          reason you should not restrict read or execute permissions
          on non-set-id executables.
        </p>

        <p>
          Some setuid programs need to be restricted to particular
          sets of users, using file permissions.  In this case they
          should be owned by the uid to which they are set-id, and by
          the group which should be allowed to execute them.  They
          should have mode 4754; again there is no point in making
          them unreadable to those users who must not be allowed to
          execute them.
        </p>

        <p>
          It is possible to arrange that the system administrator can
          reconfigure the package to correspond to their local
          security policy by changing the permissions on a binary:
          they can do this by using <prgn>dpkg-statoverride</prgn>, as
          described below.<footnote>
            Ordinary files installed by <prgn>dpkg</prgn> (as
            opposed to <tt>conffile</tt>s and other similar objects)
            normally have their permissions reset to the distributed
            permissions when the package is reinstalled.  However,
            the use of <prgn>dpkg-statoverride</prgn> overrides this
            default behavior.
          </footnote>
          Another method you should consider is to create a group for
          people allowed to use the program(s) and make any setuid
          executables executable only by that group.
        </p>

        <p>
          If you need to create a new user or group for your package
          there are two possibilities.  Firstly, you may need to
          make some files in the binary package be owned by this
          user or group, or you may need to compile the user or
          group id (rather than just the name) into the binary
          (though this latter should be avoided if possible, as in
          this case you need a statically allocated id).</p>

        <p>
          If you need a statically allocated id, you must ask for a
          user or group id from the <tt>base-passwd</tt> maintainer,
          and must not release the package until you have been
          allocated one.  Once you have been allocated one you must
          either make the package depend on a version of the
          <tt>base-passwd</tt> package with the id present in
          <file>/etc/passwd</file> or <file>/etc/group</file>, or arrange for
          your package to create the user or group itself with the
          correct id (using <tt>adduser</tt>) in its
          <prgn>preinst</prgn> or <prgn>postinst</prgn>.  (Doing it in
          the <prgn>postinst</prgn> is to be preferred if it is
          possible, otherwise a pre-dependency will be needed on the
          <tt>adduser</tt> package.)
        </p>

        <p>
          On the other hand, the program might be able to determine
          the uid or gid from the user or group name at runtime, so
          that a dynamically allocated id can be used.  In this case
          you should choose an appropriate user or group name,
          discussing this on <prgn>debian-devel</prgn> and checking
          with the <package/base-passwd/ maintainer that it is unique and that
          they do not wish you to use a statically allocated id
          instead.  When this has been checked you must arrange for
          your package to create the user or group if necessary using
          <prgn>adduser</prgn> in the <prgn>preinst</prgn> or
          <prgn>postinst</prgn> script (again, the latter is to be
          preferred if it is possible).
        </p>

        <p>
          Note that changing the numeric value of an id associated
          with a name is very difficult, and involves searching the
          file system for all appropriate files.  You need to think
          carefully whether a static or dynamic id is required, since
          changing your mind later will cause problems.
        </p>

        <sect1><heading>The use of <prgn>dpkg-statoverride</prgn></heading>
          <p>
            This section is not intended as policy, but as a
            description of the use of <prgn>dpkg-statoverride</prgn>.
          </p>

          <p>
            If a system administrator wishes to have a file (or
            directory or other such thing) installed with owner and
            permissions different from those in the distributed Debian
            package, they can use the <prgn>dpkg-statoverride</prgn>
            program to instruct <prgn>dpkg</prgn> to use the different
            settings every time the file is installed.  Thus the
            package maintainer should distribute the files with their
            normal permissions, and leave it for the system
            administrator to make any desired changes.  For example, a
            daemon which is normally required to be setuid root, but
            in certain situations could be used without being setuid,
            should be installed setuid in the <tt>.deb</tt>.  Then the
            local system administrator can change this if they wish.
            If there are two standard ways of doing it, the package
            maintainer can use <tt>debconf</tt> to find out the
            preference, and call <prgn>dpkg-statoverride</prgn> in the
            maintainer script if necessary to accommodate the system
            administrator's choice. Care must be taken during
            upgrades to not override an existing setting.
          </p>

          <p>
            Given the above, <prgn>dpkg-statoverride</prgn> is
            essentially a tool for system administrators and would not
            normally be needed in the maintainer scripts.  There is
            one type of situation, though, where calls to
            <prgn>dpkg-statoverride</prgn> would be needed in the
            maintainer scripts, and that involves packages which use
            dynamically allocated user or group ids.  In such a
            situation, something like the following idiom can be very
            helpful in the package's <prgn>postinst</prgn>, where
            <tt>sysuser</tt> is a dynamically allocated id:
            <example>
for i in /usr/bin/foo /usr/sbin/bar
do
  # only do something when no setting exists
  if ! dpkg-statoverride --list $i >/dev/null 2>&1
  then
    #include: debconf processing, question about foo and bar
    if [ "$RET" = "true" ] ; then
      dpkg-statoverride --update --add sysuser root 4755 $i
    fi
  fi
done
            </example>
            The corresponding code to remove the override when the package
            is purged would be:
            <example>
for i in /usr/bin/foo /usr/sbin/bar
do
  if dpkg-statoverride --list $i >/dev/null 2>&1
  then
    dpkg-statoverride --remove $i
  fi
done
            </example>
          </p>
        </sect1>
      </sect>

      <sect id="filenames">
        <heading>File names</heading>

        <p>
          The name of the files installed by binary packages in the system PATH
          (namely <tt>/bin</tt>, <tt>/sbin</tt>, <tt>/usr/bin</tt>,
          <tt>/usr/sbin</tt> and <tt>/usr/games</tt>) must be encoded in
          ASCII.
        </p>

        <p>
          The name of the files and directories installed by binary packages
          outside the system PATH must be encoded in UTF-8 and should be
          restricted to ASCII when it is possible to do so.
        </p>
      </sect>
    </chapt>


    <chapt id="customized-programs">
      <heading>Customized programs</heading>

      <sect id="arch-spec">
        <heading>Architecture specification strings</heading>

        <p>
          If a program needs to specify an <em>architecture specification
          string</em> in some place, it should select one of the strings
          provided by <tt>dpkg-architecture -L</tt>. The strings are in
          the format <tt><var>os</var>-<var>arch</var></tt>, though the OS
          part is sometimes elided, as when the OS is Linux.
        </p>

        <p>
          Note that we don't want to use
          <tt><var>arch</var>-debian-linux</tt> to apply to the rule
          <tt><var>architecture</var>-<var>vendor</var>-<var>os</var></tt>
          since this would make our programs incompatible with other
          Linux distributions.  We also don't use something like
          <tt><var>arch</var>-unknown-linux</tt>, since the
          <tt>unknown</tt> does not look very good.
        </p>

        <sect1 id="arch-wildcard-spec">
          <heading>Architecture wildcards</heading>

          <p>
            A package may specify an architecture wildcard. Architecture
            wildcards are in the format <tt>any</tt> (which matches every
            architecture), <tt><var>os</var></tt>-any, or
            any-<tt><var>cpu</var></tt>. <footnote>
              Internally, the package system normalizes the GNU triplets
              and the Debian arches into Debian arch triplets (which are
              kind of inverted GNU triplets), with the first component of
              the triplet representing the libc and ABI in use, and then
              does matching against those triplets.  However, such
              triplets are an internal implementation detail that should
              not be used by packages directly.  The libc and ABI portion
              is handled internally by the package system based on
              the <var>os</var> and <var>cpu</var>.
            </footnote>
          </p>
        </sect1>
      </sect>

      <sect>
        <heading>Daemons</heading>

        <p>
          The configuration files <file>/etc/services</file>,
          <file>/etc/protocols</file>, and <file>/etc/rpc</file> are managed
          by the <prgn>netbase</prgn> package and must not be modified
          by other packages.
        </p>

        <p>
          If a package requires a new entry in one of these files, the
          maintainer should get in contact with the
          <prgn>netbase</prgn> maintainer, who will add the entries
          and release a new version of the <prgn>netbase</prgn>
          package.
        </p>

        <p>
          The configuration file <file>/etc/inetd.conf</file> must not be
          modified by the package's scripts except via the
          <prgn>update-inetd</prgn> script or the
          <file>DebianNet.pm</file> Perl module.  See their documentation
          for details on how to add entries.
        </p>

        <p>
          If a package wants to install an example entry into
          <file>/etc/inetd.conf</file>, the entry must be preceded with
          exactly one hash character (<tt>#</tt>). Such lines are
          treated as "commented out by user" by the
          <prgn>update-inetd</prgn> script and are not changed or
          activated during package updates.
        </p>
      </sect>

      <sect>
        <heading>Using pseudo-ttys and modifying wtmp, utmp and
        lastlog</heading>

        <p>
          Some programs need to create pseudo-ttys. This should be done
          using Unix98 ptys if the C library supports it. The resulting
          program must not be installed setuid root, unless that
          is required for other functionality.
        </p>

        <p>
          The files <file>/var/run/utmp</file>, <file>/var/log/wtmp</file> and
          <file>/var/log/lastlog</file> must be installed writable by
          group <tt>utmp</tt>.  Programs which need to modify those
          files must be installed setgid <tt>utmp</tt>.
        </p>
      </sect>

      <sect>
        <heading>Editors and pagers</heading>

        <p>
          Some programs have the ability to launch an editor or pager
          program to edit or display a text document.  Since there are
          lots of different editors and pagers available in the Debian
          distribution, the system administrator and each user should
          have the possibility to choose their preferred editor and
          pager.
        </p>

        <p>
          In addition, every program should choose a good default
          editor/pager if none is selected by the user or system
          administrator.
        </p>

        <p>
          Thus, every program that launches an editor or pager must
          use the EDITOR or PAGER environment variable to determine
          the editor or pager the user wishes to use.  If these
          variables are not set, the programs <file>/usr/bin/editor</file>
          and <file>/usr/bin/pager</file> should be used, respectively.
        </p>

        <p>
          These two files are managed through the <prgn>dpkg</prgn>
          "alternatives" mechanism.  Every package providing an editor or
          pager must call the <prgn>update-alternatives</prgn> script to
          register as an alternative for <file>/usr/bin/editor</file>
          or <file>/usr/bin/pager</file> as appropriate.  The alternative
          should have a slave alternative
          for <file>/usr/share/man/man1/editor.1.gz</file>
          or <file>/usr/share/man/man1/pager.1.gz</file> pointing to the
          corresponding manual page.
        </p>

        <p>
          If it is very hard to adapt a program to make use of the
          EDITOR or PAGER variables, that program may be configured to
          use <file>/usr/bin/sensible-editor</file> and
          <file>/usr/bin/sensible-pager</file> as the editor or pager
          program respectively.  These are two scripts provided in the
          <package>sensible-utils</package> package that check the EDITOR
          and PAGER variables and launch the appropriate program, and fall
          back to <file>/usr/bin/editor</file>
          and <file>/usr/bin/pager</file> if the variable is not set.
        </p>

        <p>
          A program may also use the VISUAL environment variable to
          determine the user's choice of editor.  If it exists, it
          should take precedence over EDITOR.  This is in fact what
          <file>/usr/bin/sensible-editor</file> does.
        </p>

        <p>
          It is not required for a package to depend on
          <tt>editor</tt> and <tt>pager</tt>, nor is it required for a
          package to provide such virtual packages.<footnote>
              The Debian base system already provides an editor and a
              pager program.
          </footnote>
        </p>
      </sect>

      <sect id="web-appl">
        <heading>Web servers and applications</heading>

        <p>
          This section describes the locations and URLs that should
          be used by all web servers and web applications in the
          Debian system.
        </p>

        <p>
          <enumlist>
            <item>
                Cgi-bin executable files are installed in the
                directory
                <example compact="compact">
/usr/lib/cgi-bin
                </example>
                or a subdirectory of that directory, and the script
                <example compact="compact">
/usr/lib/cgi-bin/.../<var>cgi-bin-name</var>
                </example>
                should be referred to as
                <example compact="compact">
http://localhost/cgi-bin/.../<var>cgi-bin-name</var>
                </example>
            </item>

            <item>
              <p>(Deleted)</p>
            </item>

            <item>
              <p>Access to images</p>
              <p>
                It is recommended that images for a package be stored
                in <tt>/usr/share/images/<var>package</var></tt> and
                may be referred to through an alias <tt>/images/</tt>
                as
                <example>
                  http://localhost/images/&lt;package&gt;/&lt;filename&gt;     
                </example>
                
              </p>
            </item>

            <item>
              <p>Web Document Root</p>

              <p>
                Web Applications should try to avoid storing files in
                the Web Document Root.  Instead they should use the
                /usr/share/doc/<var>package</var> directory for
                documents and register the Web Application via the
                <package>doc-base</package> package.  If access to the
                web document root is unavoidable then use
                <example compact="compact">
/var/www
                </example>
                as the Document Root.  This might be just a symbolic
                link to the location where the system administrator
                has put the real document root.
              </p>
            </item>
            <item><p>Providing httpd and/or httpd-cgi</p>
              <p>
                All web servers should provide the virtual package
                <tt>httpd</tt>. If a web server has CGI support it should
                provide <tt>httpd-cgi</tt> additionally.
              </p>
              <p>
                All web applications which do not contain CGI scripts should
                depend on <tt>httpd</tt>, all those web applications which
                <tt>do</tt> contain CGI scripts, should depend on
                <tt>httpd-cgi</tt>.
              </p>
            </item>
          </enumlist>
        </p>
      </sect>

      <sect id="mail-transport-agents">
        <heading>Mail transport, delivery and user agents</heading>

        <p>
          Debian packages which process electronic mail, whether mail
          user agents (MUAs) or mail transport agents (MTAs), must
          ensure that they are compatible with the configuration
          decisions below.  Failure to do this may result in lost
          mail, broken <tt>From:</tt> lines, and other serious brain
          damage!
        </p>

        <p>
          The mail spool is <file>/var/mail</file> and the interface to
          send a mail message is <file>/usr/sbin/sendmail</file> (as per
          the FHS).  On older systems, the mail spool may be
          physically located in <file>/var/spool/mail</file>, but all
          access to the mail spool should be via the
          <file>/var/mail</file> symlink.  The mail spool is part of the
          base system and not part of the MTA package.
        </p>

        <p>
          All Debian MUAs, MTAs, MDAs and other mailbox accessing
          programs (such as IMAP daemons) must lock the mailbox in an
          NFS-safe way. This means that <tt>fcntl()</tt> locking must
          be combined with dot locking.  To avoid deadlocks, a program
          should use <tt>fcntl()</tt> first and dot locking after
          this, or alternatively implement the two locking methods in
          a non blocking way<footnote>
              If it is not possible to establish both locks, the
              system shouldn't wait for the second lock to be
              established, but remove the first lock, wait a (random)
              time, and start over locking again.
          </footnote>. Using the functions <tt>maillock</tt> and
          <tt>mailunlock</tt> provided by the
          <tt>liblockfile*</tt><footnote>
              You will need to depend on <tt>liblockfile1 (&gt;&gt;1.01)</tt>
              to use these functions.
          </footnote> packages is the recommended way to realize this.
        </p>

        <p>
          Mailboxes are generally either mode 600 and owned by
          <var>user</var> or mode 660 and owned by
          <tt><var>user</var>:mail</tt><footnote>
            There are two traditional permission schemes for mail spools:
            mode 600 with all mail delivery done by processes running as
            the destination user, or mode 660 and owned by group mail with
            mail delivery done by a process running as a system user in
            group mail.  Historically, Debian required mode 660 mail
            spools to enable the latter model, but that model has become
            increasingly uncommon and the principle of least privilege
            indicates that mail systems that use the first model should
            use permissions of 600.  If delivery to programs is permitted,
            it's easier to keep the mail system secure if the delivery
            agent runs as the destination user.  Debian Policy therefore
            permits either scheme.
          </footnote>. The local system administrator may choose a
          different permission scheme; packages should not make
          assumptions about the permission and ownership of mailboxes
          unless required (such as when creating a new mailbox).  A MUA
          may remove a mailbox (unless it has nonstandard permissions) in
          which case the MTA or another MUA must recreate it if needed.
        </p>

        <p>
          The mail spool is 2775 <tt>root:mail</tt>, and MUAs should
          be setgid mail to do the locking mentioned above (and
          must obviously avoid accessing other users' mailboxes
          using this privilege).</p>

        <p>
          <file>/etc/aliases</file> is the source file for the system mail
          aliases (e.g., postmaster, usenet, etc.), it is the one
          which the sysadmin and <prgn>postinst</prgn> scripts may
          edit.  After <file>/etc/aliases</file> is edited the program or
          human editing it must call <prgn>newaliases</prgn>.  All MTA
          packages must come with a <prgn>newaliases</prgn> program,
          even if it does nothing, but older MTA packages did not do
          this so programs should not fail if <prgn>newaliases</prgn>
          cannot be found.  Note that because of this, all MTA
          packages must have <tt>Provides</tt>, <tt>Conflicts</tt> and
          <tt>Replaces: mail-transport-agent</tt> control fields.
        </p>

        <p>
          The convention of writing <tt>forward to
            <var>address</var></tt> in the mailbox itself is not
          supported.  Use a <tt>.forward</tt> file instead.</p>

        <p>
          The <prgn>rmail</prgn> program used by UUCP
          for incoming mail should be  <file>/usr/sbin/rmail</file>.
          Likewise, <prgn>rsmtp</prgn>, for receiving
          batch-SMTP-over-UUCP, should be <file>/usr/sbin/rsmtp</file> if it
          is supported.</p>

        <p>
          If your package needs to know what hostname to use on (for
          example) outgoing news and mail messages which are generated
          locally, you should use the file <file>/etc/mailname</file>.  It
          will contain the portion after the username and <tt>@</tt>
          (at) sign for email addresses of users on the machine
          (followed by a newline).
        </p>

        <p>
          Such a package should check for the existence of this file
          when it is being configured.  If it exists, it should be
          used without comment, although an MTA's configuration script
          may wish to prompt the user even if it finds that this file
          exists.  If the file does not exist, the package should
          prompt the user for the value (preferably using
          <prgn>debconf</prgn>) and store it in <file>/etc/mailname</file>
          as well as using it in the package's configuration.  The
          prompt should make it clear that the name will not just be
          used by that package.  For example, in this situation the
          <tt>inn</tt> package could say something like:
          <example compact="compact">
Please enter the "mail name" of your system.  This is the
hostname portion of the address to be shown on outgoing
news and mail messages.  The default is
<var>syshostname</var>, your system's host name.  Mail
name ["<var>syshostname</var>"]:
          </example>
          where <var>syshostname</var> is the output of <tt>hostname
            --fqdn</tt>.
        </p>
      </sect>

      <sect>
        <heading>News system configuration</heading>

        <p>
          All the configuration files related to the NNTP (news)
          servers and clients should be located under
          <file>/etc/news</file>.</p>

        <p>
          There are some configuration issues that apply to a number
          of news clients and server packages on the machine. These
          are:

          <taglist>
            <tag><file>/etc/news/organization</file></tag>
            <item>
                A string which should appear as the
                organization header for all messages posted
                by NNTP clients on the machine
            </item>

            <tag><file>/etc/news/server</file></tag>
            <item>
                Contains the FQDN of the upstream NNTP
                server, or localhost if the local machine is
                an NNTP server.
            </item>
          </taglist>

          Other global files may be added as required for cross-package news
          configuration.
        </p>
      </sect>


      <sect>
        <heading>Programs for the X Window System</heading>

        <sect1>
          <heading>Providing X support and package priorities</heading>

          <p>
            Programs that can be configured with support for the X
            Window System must be configured to do so and must declare
            any package dependencies necessary to satisfy their
            runtime requirements when using the X Window System.  If
            such a package is of higher priority than the X packages
            on which it depends, it is required that either the
            X-specific components be split into a separate package, or
            that an alternative version of the package, which includes
            X support, be provided, or that the package's priority be
            lowered.
          </p>
        </sect1>

        <sect1>
          <heading>Packages providing an X server</heading>

          <p>
            Packages that provide an X server that, directly or
            indirectly, communicates with real input and display
            hardware should declare in their <tt>Provides</tt> control
            field that they provide the virtual
            package <tt>xserver</tt>.<footnote>
                This implements current practice, and provides an
                actual policy for usage of the <tt>xserver</tt>
                virtual package which appears in the virtual packages
                list.  In a nutshell, X servers that interface
                directly with the display and input hardware or via
                another subsystem (e.g., GGI) should provide
                <tt>xserver</tt>.  Things like <tt>Xvfb</tt>,
                <tt>Xnest</tt>, and <tt>Xprt</tt> should not.
            </footnote>
          </p>
        </sect1>

        <sect1>
          <heading>Packages providing a terminal emulator</heading>

          <p>
            Packages that provide a terminal emulator for the X Window
            System which meet the criteria listed below should declare in
            their <tt>Provides</tt> control field that they provide the
            virtual package <tt>x-terminal-emulator</tt>.  They should
            also register themselves as an alternative for
            <file>/usr/bin/x-terminal-emulator</file>, with a priority of
            20.  That alternative should have a slave alternative
            for <file>/usr/share/man/man1/x-terminal-emulator.1.gz</file>
            pointing to the corresponding manual page.
          </p>

          <p>
            To be an <tt>x-terminal-emulator</tt>, a program must:
            <list compact="compact">
              <item>
                  Be able to emulate a DEC VT100 terminal, or a
                  compatible terminal.
              </item>

              <item>
                  Support the command-line option <tt>-e
                    <var>command</var></tt>, which creates a new
                  terminal window<footnote>
                      "New terminal window" does not necessarily mean
                      a new top-level X window directly parented by
                      the window manager; it could, if the terminal
                      emulator application were so coded, be a new
                      "view" in a multiple-document interface (MDI).
                  </footnote>
                  and runs the specified <var>command</var>,
                  interpreting the entirety of the rest of the command
                  line as a command to pass straight to exec, in the
                  manner that <tt>xterm</tt> does.
              </item>

              <item>
                  Support the command-line option <tt>-T
                    <var>title</var></tt>, which creates a new terminal
                  window with the window title <var>title</var>.
              </item>
            </list>
          </p>
        </sect1>

        <sect1>
          <heading>Packages providing a window manager</heading>

          <p>
            Packages that provide a window manager should declare in
            their <tt>Provides</tt> control field that they provide the
            virtual package <tt>x-window-manager</tt>.  They should also
            register themselves as an alternative for
            <file>/usr/bin/x-window-manager</file>, with a priority
            calculated as follows:
            <list compact="compact">
              <item>
                  Start with a priority of 20.
              </item>

              <item>
                  If the window manager supports the Debian menu
                  system, add 20 points if this support is available
                  in the package's default configuration (i.e., no
                  configuration files belonging to the system or user
                  have to be edited to activate the feature); if
                  configuration files must be modified, add only 10
                  points.
                </p>
              </item>

              <item>
                  If the window manager complies with <url
                    id="http://www.freedesktop.org/wiki/Specifications/wm-spec"
                    name="The Window Manager Specification Project">,
                  written by the <url id="http://www.freedesktop.org/wiki/"
                    name="Free Desktop Group">, add 40 points.
              </item>

              <item>
                  If the window manager permits the X session to be
                  restarted using a <em>different</em> window manager
                  (without killing the X server) in its default
                  configuration, add 10 points; otherwise add none.
              </item>
            </list>
            That alternative should have a slave alternative
            for <file>/usr/share/man/man1/x-window-manager.1.gz</file>
            pointing to the corresponding manual page.
          </p>
        </sect1>

        <sect1>
          <heading>Packages providing fonts</heading>

          <p>
            Packages that provide fonts for the X Window
            System<footnote>
                For the purposes of Debian Policy, a "font for the X
                Window System" is one which is accessed via X protocol
                requests.  Fonts for the Linux console, for PostScript
                renderer, or any other purpose, do not fit this
                definition.  Any tool which makes such fonts available
                to the X Window System, however, must abide by this
                font policy.
            </footnote>
            must do a number of things to ensure that they are both
            available without modification of the X or font server
            configuration, and that they do not corrupt files used by
            other font packages to register information about
            themselves.
            <enumlist>
              <item>
                  Fonts of any type supported by the X Window System
                  must be in a separate binary package from any
                  executables, libraries, or documentation (except
                  that specific to the fonts shipped, such as their
                  license information).  If one or more of the fonts
                  so packaged are necessary for proper operation of
                  the package with which they are associated the font
                  package may be Recommended; if the fonts merely
                  provide an enhancement, a Suggests relationship may
                  be used.  Packages must not Depend on font
                  packages.<footnote>
                      This is because the X server may retrieve fonts
                      from the local file system or over the network
                      from an X font server; the Debian package system
                      is empowered to deal only with the local
                      file system.
                  </footnote>
              </item>

              <item>
                  BDF fonts must be converted to PCF fonts with the
                  <prgn>bdftopcf</prgn> utility (available in the
                  <tt>xfonts-utils</tt> package, <prgn>gzip</prgn>ped, and
                  placed in a directory that corresponds to their
                  resolution:
                  <list compact="compact">
                    <item>
                        100 dpi fonts must be placed in
                        <file>/usr/share/fonts/X11/100dpi/</file>.
                    </item>

                    <item>
                        75 dpi fonts must be placed in
                        <file>/usr/share/fonts/X11/75dpi/</file>.
                    </item>

                    <item>
                        Character-cell fonts, cursor fonts, and other
                        low-resolution fonts must be placed in
                        <file>/usr/share/fonts/X11/misc/</file>.
                    </item>
                  </list>
              </item>

              <item>
                  Type 1 fonts must be placed in
                  <file>/usr/share/fonts/X11/Type1/</file>.  If font
                  metric files are available, they must be placed here
                  as well.
              </item>

              <item>
                  Subdirectories of <file>/usr/share/fonts/X11/</file>
                  other than those listed above must be neither
                  created nor used.  (The <file>PEX</file>, <file>CID</file>,
                  <file>Speedo</file>, and <file>cyrillic</file> directories
                  are excepted for historical reasons, but installation of
                  files into these directories remains discouraged.)
              </item>

              <item>
                  Font packages may, instead of placing files directly
                  in the X font directories listed above, provide
                  symbolic links in that font directory pointing to
                  the files' actual location in the filesystem.  Such
                  a location must comply with the FHS.
              </item>

              <item>
                  Font packages should not contain both 75dpi and
                  100dpi versions of a font.  If both are available,
                  they should be provided in separate binary packages
                  with <tt>-75dpi</tt> or <tt>-100dpi</tt> appended to
                  the names of the packages containing the
                  corresponding fonts.
              </item>

              <item>
                  Fonts destined for the <file>misc</file> subdirectory
                  should not be included in the same package as 75dpi
                  or 100dpi fonts; instead, they should be provided in
                  a separate package with <tt>-misc</tt> appended to
                  its name.
              </item>

              <item>
                  Font packages must not provide the files
                  <file>fonts.dir</file>, <file>fonts.alias</file>, or
                  <file>fonts.scale</file> in a font directory:
                  <list>
                    <item>
                        <file>fonts.dir</file> files must not be provided at all.
                    </item>

                    <item>
                        <file>fonts.alias</file> and <file>fonts.scale</file>
                        files, if needed, should be provided in the
                        directory
                        <file>/etc/X11/fonts/<var>fontdir</var>/<var>package</var>.<var>extension</var></file>,
                        where <var>fontdir</var> is the name of the
                        subdirectory of
                        <file>/usr/share/fonts/X11/</file> where the
                        package's corresponding fonts are stored
                        (e.g., <tt>75dpi</tt> or <tt>misc</tt>),
                        <var>package</var> is the name of the package
                        that provides these fonts, and
                        <var>extension</var> is either <tt>scale</tt>
                        or <tt>alias</tt>, whichever corresponds to
                        the file contents.
                    </item>
                  </list>
              </item>

              <item>
                  Font packages must declare a dependency on
                  <tt>xfonts-utils</tt> in their <tt>Depends</tt>
                  or <tt>Pre-Depends</tt> control field.
              </item>

              <item>
                  Font packages that provide one or more
                  <file>fonts.scale</file> files as described above must
                  invoke <prgn>update-fonts-scale</prgn> on each
                  directory into which they installed fonts
                  <em>before</em> invoking
                  <prgn>update-fonts-dir</prgn> on that directory.
                  This invocation must occur in both the
                  <prgn>postinst</prgn> (for all arguments) and
                  <prgn>postrm</prgn> (for all arguments except
                  <tt>upgrade</tt>) scripts.
              </item>

              <item>
                  Font packages that provide one or more
                  <file>fonts.alias</file> files as described above must
                  invoke <prgn>update-fonts-alias</prgn> on each
                  directory into which they installed fonts.  This
                  invocation must occur in both the
                  <prgn>postinst</prgn> (for all arguments) and
                  <prgn>postrm</prgn> (for all arguments except
                  <tt>upgrade</tt>) scripts.
              </item>

              <item>
                  Font packages must invoke
                  <prgn>update-fonts-dir</prgn> on each directory into
                  which they installed fonts.  This invocation must
                  occur in both the <prgn>postinst</prgn> (for all
                  arguments) and <prgn>postrm</prgn> (for all
                  arguments except <tt>upgrade</tt>) scripts.
              </item>

              <item>
                  Font packages must not provide alias names for the
                  fonts they include which collide with alias names
                  already in use by fonts already packaged.
              </item>

              <item>
                  Font packages must not provide fonts with the same
                  XLFD registry name as another font already packaged.
              </item>
            </enumlist>
          </p>
        </sect1>

        <sect1 id="appdefaults">
          <heading>Application defaults files</heading>

          <p>
            Application defaults files must be installed in the
            directory <file>/etc/X11/app-defaults/</file> (use of a
            localized subdirectory of <file>/etc/X11/</file> as described
            in the <em>X Toolkit Intrinsics - C Language
            Interface</em> manual is also permitted).  They must be
            registered as <tt>conffile</tt>s or handled as
            configuration files.
          </p>

          <p>
            Customization of programs' X resources may also be
            supported with the provision of a file with the same name
            as that of the package placed in
            the <file>/etc/X11/Xresources/</file> directory, which
            must be registered as a <tt>conffile</tt> or handled as a
            configuration file.<footnote>
                Note that this mechanism is not the same as using
                app-defaults; app-defaults are tied to the client
                binary on the local file system, whereas X resources
                are stored in the X server and affect all connecting
                clients.
            </footnote>
          </p>
        </sect1>

        <sect1>
          <heading>Installation directory issues</heading>

          <p>
            Historically, packages using the X Window System used a
            separate set of installation directories from other packages.
            This practice has been discontinued and packages using the X
            Window System should now generally be installed in the same
            directories as any other package.  Specifically, packages must
            not install files under the <file>/usr/X11R6/</file> directory
            and the <file>/usr/X11R6/</file> directory hierarchy should be
            regarded as obsolete.
          </p>

          <p>
            Include files previously installed under
            <file>/usr/X11R6/include/X11/</file> should be installed into
            <file>/usr/include/X11/</file>.  For files previously
            installed into subdirectories of
            <file>/usr/X11R6/lib/X11/</file>, package maintainers should
            determine if subdirectories of <file>/usr/lib/</file> and
            <file>/usr/share/</file> can be used.  If not, a subdirectory
            of <file>/usr/lib/X11/</file> should be used.
          </p>

          <p>
            Configuration files for window, display, or session managers
            or other applications that are tightly integrated with the X
            Window System may be placed in a subdirectory
            of <file>/etc/X11/</file> corresponding to the package name.
            Other X Window System applications should use
            the <file>/etc/</file> directory unless otherwise mandated by
            policy (such as for <ref id="appdefaults">).
          </p>
        </sect1>
      </sect>

      <sect id="perl">
        <heading>Perl programs and modules</heading>

        <p>
          Perl programs and modules should follow the current Perl policy.
        </p>

        <p>
          The Perl policy can be found in the <tt>perl-policy</tt>
          files in the <tt>debian-policy</tt> package.
          It is also available from the Debian web mirrors at
          <tt><url name="/doc/packaging-manuals/perl-policy/"
                id="http://www.debian.org/doc/packaging-manuals/perl-policy/"></tt>.
        </p>
      </sect>

      <sect id="emacs">
        <heading>Emacs lisp programs</heading>

        <p>
          Please refer to the "Debian Emacs Policy" for details of how to
          package emacs lisp programs.
        </p>

        <p>
          The Emacs policy is available in
          <file>debian-emacs-policy.gz</file> of the
          <package>emacsen-common</package> package.
          It is also available from the Debian web mirrors at
          <tt><url name="/doc/packaging-manuals/debian-emacs-policy"
                id="http://www.debian.org/doc/packaging-manuals/debian-emacs-policy"></tt>.
        </p>
      </sect>

      <sect>
        <heading>Games</heading>

        <p>
          The permissions on <file>/var/games</file> are mode 755, owner
          <tt>root</tt> and group <tt>root</tt>.
        </p>

        <p>
          Each game decides on its own security policy.</p>

        <p>
          Games which require protected, privileged access to
          high-score files, saved games, etc., may be made
          set-<em>group</em>-id (mode 2755) and owned by
          <tt>root:games</tt>, and use files and directories with
          appropriate permissions (770 <tt>root:games</tt>, for
          example).  They must not be made
          set-<em>user</em>-id, as this causes security problems. (If
          an attacker can subvert any set-user-id game they can
          overwrite the executable of any other, causing other players
          of these games to run a Trojan horse program.  With a
          set-group-id game the attacker only gets access to less
          important game data, and if they can get at the other
          players' accounts at all it will take considerably more
          effort.)</p>

        <p>
          Some packages, for example some fortune cookie programs, are
          configured by the upstream authors to install with their
          data files or other static information made unreadable so
          that they can only be accessed through set-id programs
          provided.  You should not do this in a Debian package: anyone can
          download the <file>.deb</file> file and read the data from it,
          so there is no point making the files unreadable.  Not
          making the files unreadable also means that you don't have
          to make so many programs set-id, which reduces the risk of a
          security hole.</p>

        <p>
          As described in the FHS, binaries of games should be
          installed in the directory <file>/usr/games</file>. This also
          applies to games that use the X Window System. Manual pages
          for games (X and non-X games) should be installed in
          <file>/usr/share/man/man6</file>.</p>
      </sect>
    </chapt>


    <chapt id="docs">
      <heading>Documentation</heading>

      <sect>
        <heading>Manual pages</heading>

        <p>
          You should install manual pages in <prgn>nroff</prgn> source
          form, in appropriate places under <file>/usr/share/man</file>.
          You should only use sections 1 to 9 (see the FHS for more
          details).  You must not install a pre-formatted "cat page".
        </p>

        <p>
          Each program, utility, and function should have an
          associated manual page included in the same package. It is
          suggested that all configuration files also have a manual
          page included as well. Manual pages for protocols and other
          auxiliary things are optional.
        </p>

        <p>
          If no manual page is available, this is considered as a bug
          and should be reported to the Debian Bug Tracking System (the
          maintainer of the package is allowed to write this bug report
          themselves, if they so desire).  Do not close the bug report
          until a proper man page is available.<footnote>
              It is not very hard to write a man page. See the
              <url id="http://www.schweikhardt.net/man_page_howto.html"
                name="Man-Page-HOWTO">,
              <manref name="man" section="7">, the examples created
              by <prgn>dh_make</prgn>, the helper
              program <prgn>help2man</prgn>, or the
              directory <file>/usr/share/doc/man-db/examples</file>.
          </footnote>
        </p>

        <p>
          You may forward a complaint about a missing man page to the
          upstream authors, and mark the bug as forwarded in the
          Debian bug tracking system.  Even though the GNU Project do
          not in general consider the lack of a man page to be a bug,
          we do; if they tell you that they don't consider it a bug
          you should leave the bug in our bug tracking system open
          anyway.
        </p>

        <p>
          Manual pages should be installed compressed using <tt>gzip -9</tt>.
        </p>

        <p>
          If one man page needs to be accessible via several names it
          is better to use a symbolic link than the <file>.so</file>
          feature, but there is no need to fiddle with the relevant
          parts of the upstream source to change from <file>.so</file> to
          symlinks: don't do it unless it's easy.  You should not
          create hard links in the manual page directories, nor put
          absolute filenames in <file>.so</file> directives.  The filename
          in a <file>.so</file> in a man page should be relative to the
          base of the man page tree (usually
          <file>/usr/share/man</file>). If you do not create any links
          (whether symlinks, hard links, or <tt>.so</tt> directives)
          in the file system to the alternate names of the man page,
          then you should not rely on <prgn>man</prgn> finding your
          man page under those names based solely on the information in
          the man page's header.<footnote>
              Supporting this in <prgn>man</prgn> often requires
              unreasonable processing time to find a manual page or to
              report that none exists, and moves knowledge into man's
              database that would be better left in the file system.
              This support is therefore deprecated and will cease to
              be present in the future.
          </footnote>
        </p>

        <p>
          Manual pages in locale-specific subdirectories of
          <file>/usr/share/man</file> should use either UTF-8 or the usual
          legacy encoding for that language (normally the one corresponding
          to the shortest relevant locale name in
          <file>/usr/share/i18n/SUPPORTED</file>). For example, pages under
          <file>/usr/share/man/fr</file> should use either UTF-8 or
          ISO-8859-1.<footnote>
            <prgn>man</prgn> will automatically detect whether UTF-8 is in
            use. In future, all manual pages will be required to use
            UTF-8.
          </footnote>
        </p>

        <p>
          A country name (the <tt>DE</tt> in <tt>de_DE</tt>) should not be
          included in the subdirectory name unless it indicates a
          significant difference in the language, as this excludes
          speakers of the language in other countries.<footnote>
            At the time of writing, Chinese and Portuguese are the main
            languages with such differences, so <file>pt_BR</file>,
            <file>zh_CN</file>, and <file>zh_TW</file> are all allowed.
          </footnote>
        </p>

        <p>
          If a localized version of a manual page is provided, it should
          either be up-to-date or it should be obvious to the reader that
          it is outdated and the original manual page should be used
          instead.  This can be done either by a note at the beginning of
          the manual page or by showing the missing or changed portions in
          the original language instead of the target language.
        </p>
      </sect>

      <sect>
        <heading>Info documents</heading>

        <p>
          Info documents should be installed in <file>/usr/share/info</file>.
          They should be compressed with <tt>gzip -9</tt>.
        </p>

        <p>
          The <prgn>install-info</prgn> program maintains a directory of
          installed info documents in <file>/usr/share/info/dir</file> for the
          use of info readers.  This file must not be included in packages
          other than <package>install-info</package>.
        </p>

        <p>
          <prgn>install-info</prgn> is automatically invoked when
          appropriate using dpkg triggers.  Packages other than
          <package>install-info</package> <em>should not</em> invoke
          <prgn>install-info</prgn> directly and <em>should not</em>
          depend on, recommend, or suggest <package>install-info</package>
          for this purpose.
        </p>

        <p>
          Info readers requiring the <file>/usr/share/info/dir</file> file
          should depend on <package>install-info</package>.
        </p>

        <p>
          Info documents should contain section and directory entry
          information in the document for the use
          of <prgn>install-info</prgn>.  The section should be specified
          via a line starting with <tt>INFO-DIR-SECTION</tt> followed by a
          space and the section of this info page.  The directory entry or
          entries should be included between
          a <tt>START-INFO-DIR-ENTRY</tt> line and
          an <tt>END-INFO-DIR-ENTRY</tt> line.  For example:
          <example>
INFO-DIR-SECTION Individual utilities
START-INFO-DIR-ENTRY
* example: (example).               An example info directory entry.
END-INFO-DIR-ENTRY
          </example>
          To determine which section to use, you should look
          at <file>/usr/share/info/dir</file> on your system and choose
          the most relevant (or create a new section if none of the
          current sections are relevant).<footnote>
            Normally, info documents are generated from Texinfo source.
            To include this information in the generated info document, if
            it is absent, add commands like:
            <example>
@dircategory Individual utilities
@direntry
* example: (example).               An example info directory entry.
@end direntry
            </example>
            to the Texinfo source of the document and ensure that the info
            documents are rebuilt from source during the package build.
          </footnote>
        </p>
      </sect>

      <sect>
        <heading>Additional documentation</heading>

        <p>
          Any additional documentation that comes with the package may
          be installed at the discretion of the package maintainer.
          Plain text documentation should be installed in the directory
          <file>/usr/share/doc/<var>package</var></file>, where
          <var>package</var> is the name of the package, and
          compressed with <tt>gzip -9</tt> unless it is small.
        </p>

        <p>
          If a package comes with large amounts of documentation which
          many users of the package will not require you should create
          a separate binary package to contain it, so that it does not
          take up disk space on the machines of users who do not need
          or want it installed.</p>

        <p>
          It is often a good idea to put text information files
          (<file>README</file>s, changelogs, and so forth) that come with
          the source package in <file>/usr/share/doc/<var>package</var></file>
          in the binary package.  However, you don't need to install
          the instructions for building and installing the package, of
          course!</p>

        <p>
          Packages must not require the existence of any files in
          <file>/usr/share/doc/</file> in order to function
          <footnote>
              The system administrator should be able to
              delete files in <file>/usr/share/doc/</file> without causing
              any programs to break.
          </footnote>.
          Any files that are referenced by programs but are also
          useful as stand alone documentation should be installed under
          <file>/usr/share/<var>package</var>/</file> with symbolic links from
          <file>/usr/share/doc/<var>package</var></file>.
        </p>

        <p>
          <file>/usr/share/doc/<var>package</var></file> may be a symbolic
          link to another directory in <file>/usr/share/doc</file> only if
          the two packages both come from the same source and the
          first package Depends on the second.<footnote>
            <p>
              Please note that this does not override the section on
              changelog files below, so the file 
              <file>/usr/share/doc/<var>package</var>/changelog.Debian.gz</file>
              must refer to the changelog for the current version of
              <var>package</var> in question. In practice, this means
              that the sources of the target and the destination of the
              symlink must be the same (same source package and
              version). 
            </p>
          </footnote>
        </p>

        <p>
          Former Debian releases placed all additional documentation
          in <file>/usr/doc/<var>package</var></file>.  This has been
          changed to <file>/usr/share/doc/<var>package</var></file>,
          and packages must not put documentation in the directory
          <file>/usr/doc/<var>package</var></file>. <footnote>
            At this phase of the transition, we no longer require a
            symbolic link in <file>/usr/doc/</file>. At a later point,
            policy shall change to make the symbolic links a bug.
          </footnote>
        </p>
      </sect>

      <sect>
        <heading>Preferred documentation formats</heading>

        <p>
          The unification of Debian documentation is being carried out
          via HTML.</p>

        <p>
          If your package comes with extensive documentation in a
          markup format that can be converted to various other formats
          you should if possible ship HTML versions in a binary
          package, in the directory
          <file>/usr/share/doc/<var>appropriate-package</var></file> or
          its subdirectories.<footnote>
              The rationale: The important thing here is that HTML
              docs should be available in <em>some</em> package, not
              necessarily in the main binary package.
          </footnote>
        </p>

        <p>
          Other formats such as PostScript may be provided at the
          package maintainer's discretion.
        </p>
      </sect>

      <sect id="copyrightfile">
        <heading>Copyright information</heading>

        <p>
          Every package must be accompanied by a verbatim copy of its
          copyright information and distribution license in the file
          <file>/usr/share/doc/<var>package</var>/copyright</file>. This
          file must neither be compressed nor be a symbolic link.
        </p>

        <p>
          In addition, the copyright file must say where the upstream
          sources (if any) were obtained, and should name the original
          authors.
        </p>

        <p>
          Packages in the <em>contrib</em> or <em>non-free</em> archive
          areas should state in the copyright file that the package is not
          part of the Debian distribution and briefly explain why.
        </p>

        <p>
          A copy of the file which will be installed in
          <file>/usr/share/doc/<var>package</var>/copyright</file> should
          be in <file>debian/copyright</file> in the source package.
        </p>

        <p>
          <file>/usr/share/doc/<var>package</var></file> may be a symbolic
          link to another directory in <file>/usr/share/doc</file> only if
          the two packages both come from the same source and the
          first package Depends on the second.  These rules are important
          because <file>copyright</file> files must be extractable by
          mechanical means.
        </p>

        <p>
          Packages distributed under the Apache license (version 2.0), the
          Artistic license, the GNU GPL (versions 1, 2, or 3), the GNU
          LGPL (versions 2, 2.1, or 3), and the GNU FDL (versions 1.2 or
          1.3) should refer to the corresponding files
          under <file>/usr/share/common-licenses</file>,<footnote>
            <p>
              In particular,
              <file>/usr/share/common-licenses/Apache-2.0</file>,
              <file>/usr/share/common-licenses/Artistic</file>,
              <file>/usr/share/common-licenses/GPL-1</file>,
              <file>/usr/share/common-licenses/GPL-2</file>,
              <file>/usr/share/common-licenses/GPL-3</file>,
              <file>/usr/share/common-licenses/LGPL-2</file>,
              <file>/usr/share/common-licenses/LGPL-2.1</file>,
              <file>/usr/share/common-licenses/LGPL-3</file>,
              <file>/usr/share/common-licenses/GFDL-1.2</file>, and
              <file>/usr/share/common-licenses/GFDL-1.3</file>
              respectively.  The University of California BSD license is
              also included in <package>base-files</package> as
              <file>/usr/share/common-licenses/BSD</file>, but given the
              brevity of this license, its specificity to code whose
              copyright is held by the Regents of the University of
              California, and the frequency of minor wording changes, its
              text should be included in the copyright file rather than
              referencing this file.
            </p>
          </footnote> rather than quoting them in the copyright
          file. 
        </p>

        <p>
          You should not use the copyright file as a general <file>README</file>
          file.  If your package has such a file it should be
          installed in <file>/usr/share/doc/<var>package</var>/README</file> or
          <file>README.Debian</file> or some other appropriate place.
        </p>

        <p>
          All copyright files must be encoded in UTF-8.
        </p>

        <sect1 id="copyrightformat">
          <heading>Machine-readable copyright information</heading>

          <p>
            A specification for a standard, machine-readable format
            for <file>debian/copyright</file> files is maintained as part
            of the <package>debian-policy</package> package.  This
            document may be found in the <file>copyright-format</file>
            files in the <package>debian-policy</package> package.  It is
            also available from the Debian web mirrors at
            <tt><url name="/doc/packaging-manuals/copyright-format/1.0/"
                     id="http://www.debian.org/doc/packaging-manuals/copyright-format/1.0/"></tt>.
          </p>

          <p>
            Use of this format is optional.
          </p>
        </sect1>
      </sect>

      <sect>
        <heading>Examples</heading>

        <p>
          Any examples (configurations, source files, whatever),
          should be installed in a directory
          <file>/usr/share/doc/<var>package</var>/examples</file>.  These
          files should not be referenced by any program: they're there
          for the benefit of the system administrator and users as
          documentation only.  Architecture-specific example files
          should be installed in a directory
          <file>/usr/lib/<var>package</var>/examples</file> with symbolic
          links to them from
          <file>/usr/share/doc/<var>package</var>/examples</file>, or the
          latter directory itself may be a symbolic link to the
          former.
        </p>

        <p>
          If the purpose of a package is to provide examples, then the
          example files may be installed into
          <file>/usr/share/doc/<var>package</var></file>.
        </p>
      </sect>

      <sect id="changelogs">
        <heading>Changelog files</heading>

        <p>
          Packages that are not Debian-native must contain a
          compressed copy of the <file>debian/changelog</file> file from
          the Debian source tree in
          <file>/usr/share/doc/<var>package</var></file> with the name
          <file>changelog.Debian.gz</file>.
        </p>

        <p>
          If an upstream changelog is available, it should be accessible as
          <file>/usr/share/doc/<var>package</var>/changelog.gz</file> in
          plain text.  If the upstream changelog is distributed in
          HTML, it should be made available in that form as
          <file>/usr/share/doc/<var>package</var>/changelog.html.gz</file>
          and a plain text <file>changelog.gz</file> should be generated
          from it using, for example, <tt>lynx -dump -nolist</tt>.  If
          the upstream changelog files do not already conform to this
          naming convention, then this may be achieved either by
          renaming the files, or by adding a symbolic link, at the
          maintainer's discretion.<footnote>
              Rationale: People should not have to look in places for
              upstream changelogs merely because they are given
              different names or are distributed in HTML format.
          </footnote>
        </p>

        <p>
          All of these files should be installed compressed using
          <tt>gzip -9</tt>, as they will become large with time even
          if they start out small.
        </p>

        <p>
          If the package has only one changelog which is used both as
          the Debian changelog and the upstream one because there is
          no separate upstream maintainer then that changelog should
          usually be installed as
          <file>/usr/share/doc/<var>package</var>/changelog.gz</file>; if
          there is a separate upstream maintainer, but no upstream
          changelog, then the Debian changelog should still be called
          <file>changelog.Debian.gz</file>.
        </p>

        <p>
          For details about the format and contents of the Debian
          changelog file, please see <ref id="dpkgchangelog">.
        </p>
      </sect>
    </chapt>

    <appendix id="pkg-scope">
      <heading>Introduction and scope of these appendices</heading>

      <p>
        These appendices are taken essentially verbatim from the
        now-deprecated Packaging Manual, version 3.2.1.0.  They are
        the chapters which are likely to be of use to package
        maintainers and which have not already been included in the
        policy document itself. Most of these sections are very likely
        not relevant to policy; they should be treated as
        documentation for the packaging system. Please note that these
        appendices are included for convenience, and for historical
        reasons: they used to be part of policy package, and they have
        not yet been incorporated into dpkg documentation. However,
        they still have value, and hence they are presented here.
      </p>

      <p>
        They have not yet been checked to ensure that they are
        compatible with the contents of policy, and if there are any
        contradictions, the version in the main policy document takes
        precedence.  The remaining chapters of the old Packaging
        Manual have also not been read in detail to ensure that there
        are not parts which have been left out.  Both of these will be
        done in due course.
      </p>

      <p>
        Certain parts of the Packaging manual were integrated into the
        Policy Manual proper, and removed from the appendices. Links
        have been placed from the old locations to the new ones.
      </p>

      <p>
        <prgn>dpkg</prgn> is a suite of programs for creating binary
        package files and installing and removing them on Unix
        systems.<footnote>
            <prgn>dpkg</prgn> is targeted primarily at Debian, but may
            work on or be ported to other systems.
        </footnote>
      </p>

      <p>
        The binary packages are designed for the management of
        installed executable programs (usually compiled binaries) and
        their associated data, though source code examples and
        documentation are provided as part of some packages.</p>

      <p>
        This manual describes the technical aspects of creating Debian
        binary packages (<file>.deb</file> files).  It documents the
        behavior of the package management programs
        <prgn>dpkg</prgn>, <prgn>dselect</prgn> et al. and the way
        they interact with packages.</p>

      <p>
        This manual does not go into detail about the options and
        usage of the package building and installation tools.  It
        should therefore be read in conjunction with those programs'
        man pages.
      </p>

      <p>
        The utility programs which are provided with <prgn>dpkg</prgn>
        not described in detail here, are documented in their man pages.
      </p>

      <p>
        It is assumed that the reader is reasonably familiar with the
        <prgn>dpkg</prgn> System Administrators' manual.
        Unfortunately this manual does not yet exist.
      </p>

      <p>
        The Debian version of the FSF's GNU hello program is provided as
        an example for people wishing to create Debian packages. However,
        while the examples are helpful, they do not replace the need to
        read and follow the Policy and Programmer's Manual.</p>
    </appendix>

    <appendix id="pkg-binarypkg">
      <heading>Binary packages (from old Packaging Manual)</heading>

      <p>
        See <manref name="deb" section="5"> and <ref id="pkg-controlarea">.
      </p>

      <sect id="pkg-bincreating"><heading>Creating package files -
      <prgn>dpkg-deb</prgn>
        </heading>

        <p>
          All manipulation of binary package files is done by
          <prgn>dpkg-deb</prgn>; it's the only program that has
          knowledge of the format.  (<prgn>dpkg-deb</prgn> may be
          invoked by calling <prgn>dpkg</prgn>, as <prgn>dpkg</prgn>
          will spot that the options requested are appropriate to
          <prgn>dpkg-deb</prgn> and invoke that instead with the same
          arguments.)
        </p>

        <p>
          In order to create a binary package you must make a
          directory tree which contains all the files and directories
          you want to have in the file system data part of the package.
          In Debian-format source packages this directory is usually
          <file>debian/tmp</file>, relative to the top of the package's
          source tree.
        </p>

        <p>
          They should have the locations (relative to the root of the
          directory tree you're constructing) ownerships and
          permissions which you want them to have on the system when
          they are installed.
        </p>

        <p>
          With current versions of <prgn>dpkg</prgn> the uid/username
          and gid/groupname mappings for the users and groups being
          used should be the same on the system where the package is
          built and the one where it is installed.
        </p>

        <p>
          You need to add one special directory to the root of the
          miniature file system tree you're creating:
          <prgn>DEBIAN</prgn>.  It should contain the control
          information files, notably the binary package control file
          (see <ref id="pkg-controlfile">).
        </p>

        <p>
          The <prgn>DEBIAN</prgn> directory will not appear in the
          file system archive of the package, and so won't be installed
          by <prgn>dpkg</prgn> when the package is unpacked.
        </p>

        <p>
          When you've prepared the package, you should invoke:
          <example>
  dpkg --build <var>directory</var>
          </example>
        </p>

        <p>
          This will build the package in
          <file><var>directory</var>.deb</file>.  (<prgn>dpkg</prgn> knows
          that <tt>--build</tt> is a <prgn>dpkg-deb</prgn> option, so
          it invokes <prgn>dpkg-deb</prgn> with the same arguments to
          build the package.)
        </p>

        <p>
          See the man page <manref name="dpkg-deb" section="8"> for details of how
          to examine the contents of this newly-created file.  You may find the
          output of following commands enlightening:
          <example>
  dpkg-deb --info <var>filename</var>.deb
  dpkg-deb --contents <var>filename</var>.deb
  dpkg --contents <var>filename</var>.deb
          </example>
          To view the copyright file for a package you could use this command:
          <example>
  dpkg --fsys-tarfile <var>filename</var>.deb | tar xOf - --wildcards \*/copyright | pager
          </example>
        </p>
      </sect>

      <sect id="pkg-controlarea">
        <heading>Package control information files</heading>

        <p>
          The control information portion of a binary package is a
          collection of files with names known to <prgn>dpkg</prgn>.
          It will treat the contents of these files specially - some
          of them contain information used by <prgn>dpkg</prgn> when
          installing or removing the package; others are scripts which
          the package maintainer wants <prgn>dpkg</prgn> to run.
        </p>

        <p>
          It is possible to put other files in the package control
          information file area, but this is not generally a good idea
          (though they will largely be ignored).
        </p>

        <p>
          Here is a brief list of the control information files supported
          by <prgn>dpkg</prgn> and a summary of what they're used for.
        </p>

        <p>
          <taglist>
            <tag><tt>control</tt>
            <item>
              <p>
                This is the key description file used by
                <prgn>dpkg</prgn>.  It specifies the package's name
                and version, gives its description for the user,
                states its relationships with other packages, and so
                forth.  See <ref id="sourcecontrolfiles"> and
                <ref id="binarycontrolfiles">.
              </p>

              <p>
                It is usually generated automatically from information
                in the source package by the
                <prgn>dpkg-gencontrol</prgn> program, and with
                assistance from <prgn>dpkg-shlibdeps</prgn>.
                See <ref id="pkg-sourcetools">.
              </p>
            </item>

            <tag><tt>postinst</tt>, <tt>preinst</tt>, <tt>postrm</tt>,
                 <tt>prerm</tt>
            </tag>
            <item>
              <p>
                These are executable files (usually scripts) which
                <prgn>dpkg</prgn> runs during installation, upgrade
                and removal of packages.  They allow the package to
                deal with matters which are particular to that package
                or require more complicated processing than that
                provided by <prgn>dpkg</prgn>.  Details of when and
                how they are called are in <ref id="maintainerscripts">.
              </p>

              <p>
                It is very important to make these scripts idempotent.
                See <ref id="idempotency">.
              </p>

              <p>
                The maintainer scripts are not guaranteed to run with a
                controlling terminal and may not be able to interact with
                the user.  See <ref id="controllingterminal">.
              </p>
            </item>

            <tag><tt>conffiles</tt>
            </tag>
            <item>
                This file contains a list of configuration files which
                are to be handled automatically by <prgn>dpkg</prgn>
                (see <ref id="pkg-conffiles">).  Note that not necessarily
                every configuration file should be listed here.
            </item>

            <tag><tt>shlibs</tt>
            </tag>
            <item>
                This file contains a list of the shared libraries
                supplied by the package, with dependency details for
                each.  This is used by <prgn>dpkg-shlibdeps</prgn>
                when it determines what dependencies are required in a
                package control file. The <tt>shlibs</tt> file format
                is described on <ref id="shlibs">.
            </item>
          </taglist>
        </p>

      <sect id="pkg-controlfile">
        <heading>The main control information file: <tt>control</tt></heading>

        <p>
          The most important control information file used by
          <prgn>dpkg</prgn> when it installs a package is
          <tt>control</tt>.  It contains all the package's "vital
          statistics".
        </p>

        <p>
          The binary package control files of packages built from
          Debian sources are made by a special tool,
          <prgn>dpkg-gencontrol</prgn>, which reads
          <file>debian/control</file> and <file>debian/changelog</file> to
          find the information it needs.  See <ref id="pkg-sourcepkg"> for
          more details.
        </p>

        <p>
          The fields in binary package control files are listed in
          <ref id="binarycontrolfiles">.
        </p>

        <p>
          A description of the syntax of control files and the purpose
          of the fields is available in <ref id="controlfields">.
        </p>
      </sect>

      <sect>
        <heading>Time Stamps</heading>

        <p>
          See <ref id="timestamps">.
        </p>
      </sect>
    </appendix>

    <appendix id="pkg-sourcepkg">
      <heading>Source packages (from old Packaging Manual) </heading>

      <p>
        The Debian binary packages in the distribution are generated
        from Debian sources, which are in a special format to assist
        the easy and automatic building of binaries.
      </p>

      <sect id="pkg-sourcetools">
        <heading>Tools for processing source packages</heading>

        <p>
          Various tools are provided for manipulating source packages;
          they pack and unpack sources and help build of binary
          packages and help manage the distribution of new versions.
        </p>

        <p>
          They are introduced and typical uses described here; see
          <manref name="dpkg-source" section="1"> for full
          documentation about their arguments and operation.
        </p>

        <p>
          For examples of how to construct a Debian source package,
          and how to use those utilities that are used by Debian
          source packages, please see the <prgn>hello</prgn> example
          package.
        </p>

        <sect1 id="pkg-dpkg-source">
          <heading>
            <prgn>dpkg-source</prgn> - packs and unpacks Debian source
            packages
          </heading>

          <p>
            This program is frequently used by hand, and is also
            called from package-independent automated building scripts
            such as <prgn>dpkg-buildpackage</prgn>.
          </p>

          <p>
            To unpack a package it is typically invoked with
            <example>
  dpkg-source -x <var>.../path/to/filename</var>.dsc
            </example>
          </p>

           <p>
            with the <file><var>filename</var>.tar.gz</file> and
            <file><var>filename</var>.diff.gz</file> (if applicable) in
            the same directory.  It unpacks into
            <file><var>package</var>-<var>version</var></file>, and if
            applicable
            <file><var>package</var>-<var>version</var>.orig</file>, in
            the current directory.
          </p>

          <p>
            To create a packed source archive it is typically invoked:
            <example>
  dpkg-source -b <var>package</var>-<var>version</var>
          </example>
          </p>

          <p>
            This will create the <file>.dsc</file>, <file>.tar.gz</file> and
            <file>.diff.gz</file> (if appropriate) in the current
            directory.  <prgn>dpkg-source</prgn> does not clean the
            source tree first - this must be done separately if it is
            required.
          </p>

          <p>
            See also <ref id="pkg-sourcearchives">.</p>
        </sect1>


        <sect1 id="pkg-dpkg-buildpackage">
          <heading>
            <prgn>dpkg-buildpackage</prgn> - overall package-building
            control script
          </heading>

          <p>
            See <manref name="dpkg-buildpackage" section="1">.
          </p>
        </sect1>

        <sect1 id="pkg-dpkg-gencontrol">
          <heading>
            <prgn>dpkg-gencontrol</prgn> - generates binary package
            control files
          </heading>

          <p>
            This program is usually called from <file>debian/rules</file>
            (see <ref id="pkg-sourcetree">) in the top level of the source
            tree.
          </p>

          <p>
            This is usually done just before the files and directories in the
            temporary directory tree where the package is being built have their
            permissions and ownerships set and the package is constructed using
            <prgn>dpkg-deb/</prgn>
              <footnote>
                This is so that the control file which is produced has
                the right permissions
            </footnote>.
          </p>

          <p>
            <prgn>dpkg-gencontrol</prgn> must be called after all the
            files which are to go into the package have been placed in
            the temporary build directory, so that its calculation of
            the installed size of a package is correct.
          </p>

          <p>
            It is also necessary for <prgn>dpkg-gencontrol</prgn> to
            be run after <prgn>dpkg-shlibdeps</prgn> so that the
            variable substitutions created by
            <prgn>dpkg-shlibdeps</prgn> in <file>debian/substvars</file>
            are available.
          </p>

          <p>
            For a package which generates only one binary package, and
            which builds it in <file>debian/tmp</file> relative to the top
            of the source package, it is usually sufficient to call
            <prgn>dpkg-gencontrol</prgn>.
          </p>

          <p>
            Sources which build several binaries will typically need
            something like:
            <example>
  dpkg-gencontrol -Pdebian/tmp-<var>pkg</var> -p<var>package</var>
            </example> The <tt>-P</tt> tells
            <prgn>dpkg-gencontrol</prgn> that the package is being
            built in a non-default directory, and the <tt>-p</tt>
            tells it which package's control file should be generated.
          </p>

          <p>
            <prgn>dpkg-gencontrol</prgn> also adds information to the
            list of files in <file>debian/files</file>, for the benefit of
            (for example) a future invocation of
            <prgn>dpkg-genchanges</prgn>.</p>
        </sect1>

        <sect1 id="pkg-dpkg-shlibdeps">
          <heading>
            <prgn>dpkg-shlibdeps</prgn> - calculates shared library
            dependencies
          </heading>

          <p>
            See <manref name="dpkg-shlibdeps" section="1">.
          </p>
        </sect1>

        <sect1 id="pkg-dpkg-distaddfile">
          <heading>
            <prgn>dpkg-distaddfile</prgn> - adds a file to
            <file>debian/files</file>
          </heading>

          <p>
            Some packages' uploads need to include files other than
            the source and binary package files.
          </p>

          <p>
            <prgn>dpkg-distaddfile</prgn> adds a file to the
            <file>debian/files</file> file so that it will be included in
            the <file>.changes</file> file when
            <prgn>dpkg-genchanges</prgn> is run.
          </p>

          <p>
            It is usually invoked from the <tt>binary</tt> target of
            <file>debian/rules</file>:
            <example>
  dpkg-distaddfile <var>filename</var> <var>section</var> <var>priority</var>
            </example>
            The <var>filename</var> is relative to the directory where
            <prgn>dpkg-genchanges</prgn> will expect to find it - this
            is usually the directory above the top level of the source
            tree.  The <file>debian/rules</file> target should put the
            file there just before or just after calling
            <prgn>dpkg-distaddfile</prgn>.
          </p>

          <p>
            The <var>section</var> and <var>priority</var> are passed
            unchanged into the resulting <file>.changes</file> file.
          </p>
        </sect1>


        <sect1 id="pkg-dpkg-genchanges">
          <heading>
            <prgn>dpkg-genchanges</prgn> - generates a <file>.changes</file>
            upload control file
          </heading>

          <p>
            See <manref name="dpkg-genchanges" section="1">.
          </p>
        </sect1>

        <sect1 id="pkg-dpkg-parsechangelog">
          <heading>
            <prgn>dpkg-parsechangelog</prgn> - produces parsed
            representation of a changelog
          </heading>

          <p>
            See <manref name="dpkg-parsechangelog" section="1">.
          </p>
        </sect1>

        <sect1 id="pkg-dpkg-architecture">
          <heading>
            <prgn>dpkg-architecture</prgn> - information about the build and
            host system
          </heading>

          <p>
            See <manref name="dpkg-architecture" section="1">.
          </p>
        </sect1>
      </sect>

      <sect id="pkg-sourcetree">
        <heading>The Debian package source tree</heading>

        <p>
          The source archive scheme described later is intended to
          allow a Debian package source tree with some associated
          control information to be reproduced and transported easily.
          The Debian package source tree is a version of the original
          program with certain files added for the benefit of the
          packaging process, and with any other changes required
          made to the rest of the source code and installation
          scripts.
        </p>

        <p>
          The extra files created for Debian are in the subdirectory
          <file>debian</file> of the top level of the Debian package
          source tree. They are described below.
        </p>

        <sect1 id="pkg-debianrules">
          <heading><file>debian/rules</file> - the main building script</heading>

          <p>
            See <ref id="debianrules">.
          </p>
        </sect1>

        <sect1 id="pkg-srcsubstvars">
          <heading><file>debian/substvars</file> and variable substitutions</heading>

          <p>
            See <ref id="substvars">.
          </p>

        </sect1>

        <sect1>
          <heading><file>debian/files</file></heading>

          <p>
            See <ref id="debianfiles">.
          </p>
        </sect1>

        <sect1><heading><file>debian/tmp</file>
          </heading>

          <p>
            This is the canonical temporary location for the
            construction of binary packages by the <tt>binary</tt>
            target.  The directory <file>tmp</file> serves as the root of
            the file system tree as it is being constructed (for
            example, by using the package's upstream makefiles install
            targets and redirecting the output there), and it also
            contains the <tt>DEBIAN</tt> subdirectory.  See <ref
            id="pkg-bincreating">.
          </p>

          <p>
            If several binary packages are generated from the same
            source tree it is usual to use several
            <file>debian/tmp<var>something</var></file> directories, for
            example <file>tmp-a</file> or <file>tmp-doc</file>.
          </p>

          <p>
            Whatever <file>tmp</file> directories are created and used by
            <tt>binary</tt> must of course be removed by the
            <tt>clean</tt> target.</p></sect1>
      </sect>


      <sect id="pkg-sourcearchives"><heading>Source packages as archives
        </heading>

        <p>
          As it exists on the FTP site, a Debian source package
          consists of three related files.  You must have the right
          versions of all three to be able to use them.
        </p>

        <p>
          <taglist>
            <tag>Debian source control file - <tt>.dsc</tt></tag>
            <item>
                This file is a control file used by <prgn>dpkg-source</prgn>
                to extract a source package.
                See <ref id="debiansourcecontrolfiles">.
            </item>

            <tag>
              Original source archive -
              <file>
                <var>package</var>_<var>upstream-version</var>.orig.tar.gz
              </file>
            </tag>

            <item>
              <p>
                This is a compressed (with <tt>gzip -9</tt>)
                <prgn>tar</prgn> file containing the source code from
                the upstream authors of the program.
              </p>
            </item>

            <tag>
              Debian package diff -
              <file>
                <var>package</var>_<var>upstream_version-revision</var>.diff.gz
              </file>
            </tag>
            <item>

              <p>
                This is a unified context diff (<tt>diff -u</tt>)
                giving the changes which are required to turn the
                original source into the Debian source.  These changes
                may only include editing and creating plain files.
                The permissions of files, the targets of symbolic
                links and the characteristics of special files or
                pipes may not be changed and no files may be removed
                or renamed.
              </p>

              <p>
                All the directories in the diff must exist, except the
                <file>debian</file> subdirectory of the top of the source
                tree, which will be created by
                <prgn>dpkg-source</prgn> if necessary when unpacking.
              </p>

              <p>
                The <prgn>dpkg-source</prgn> program will
                automatically make the <file>debian/rules</file> file
                executable (see below).</p></item>
          </taglist>
        </p>

        <p>
          If there is no original source code - for example, if the
          package is specially prepared for Debian or the Debian
          maintainer is the same as the upstream maintainer - the
          format is slightly different: then there is no diff, and the
          tarfile is named
          <file><var>package</var>_<var>version</var>.tar.gz</file>,
          and preferably contains a directory named
          <file><var>package</var>-<var>version</var></file>.
        </p>
      </sect>

      <sect>
        <heading>Unpacking a Debian source package without <prgn>dpkg-source</prgn></heading>

        <p>
          <tt>dpkg-source -x</tt> is the recommended way to unpack a
          Debian source package.  However, if it is not available it
          is possible to unpack a Debian source archive as follows:
        <enumlist compact="compact">
          <item>
            <p>
              Untar the tarfile, which will create a <file>.orig</file>
              directory.</p>
          </item>
          <item>
            <p>Rename the <file>.orig</file> directory to
              <file><var>package</var>-<var>version</var></file>.</p>
          </item>
            <item>
            <p>
              Create the subdirectory <file>debian</file> at the top of
              the source tree.</p>
          </item>
          <item><p>Apply the diff using <tt>patch -p0</tt>.</p>
          </item>
          <item><p>Untar the tarfile again if you want a copy of the original
              source code alongside the Debian version.</p>
          </item>
        </enumlist>

        <p>
          It is not possible to generate a valid Debian source archive
          without using <prgn>dpkg-source</prgn>.  In particular,
          attempting to use <prgn>diff</prgn> directly to generate the
          <file>.diff.gz</file> file will not work.
        </p>

        <sect1>
          <heading>Restrictions on objects in source packages</heading>

          <p>
            The source package may not contain any hard links
            <footnote>
                This is not currently detected when building source
                packages, but only when extracting
                them.
            </footnote>
            <footnote>
                Hard links may be permitted at some point in the
                future, but would require a fair amount of
                work.
            </footnote>, device special files, sockets or setuid or
            setgid files.
            <footnote>
                Setgid directories are allowed.
            </footnote>
          </p>

          <p>
            The source packaging tools manage the changes between the
            original and Debian source using <prgn>diff</prgn> and
            <prgn>patch</prgn>.  Turning the original source tree as
            included in the <file>.orig.tar.gz</file> into the Debian
            package source must not involve any changes which cannot be
            handled by these tools.  Problematic changes which cause
            <prgn>dpkg-source</prgn> to halt with an error when
            building the source package are:
            <list compact="compact">
              <item><p>Adding or removing symbolic links, sockets or pipes.</p>
              </item>
              <item><p>Changing the targets of symbolic links.</p>
              </item>
              <item><p>Creating directories, other than <file>debian</file>.</p>
              </item>
              <item><p>Changes to the contents of binary files.</p></item>
            </list> Changes which cause <prgn>dpkg-source</prgn> to
            print a warning but continue anyway are:
            <list compact="compact">
              <item>
                <p>
                  Removing files, directories or symlinks.
                  <footnote>
                      Renaming a file is not treated specially - it is
                      seen as the removal of the old file (which
                      generates a warning, but is otherwise ignored),
                      and the creation of the new one.
                  </footnote>
                </p>
              </item>
              <item>
                <p>
                  Changed text files which are missing the usual final
                  newline (either in the original or the modified
                  source tree).
                </p>
              </item>
            </list>
            Changes which are not represented, but which are not detected by
            <prgn>dpkg-source</prgn>, are:
            <list compact="compact">
              <item><p>Changing the permissions of files (other than
                  <file>debian/rules</file>) and directories.</p></item>
            </list>
          </p>

          <p>
            The <file>debian</file> directory and <file>debian/rules</file>
            are handled specially by <prgn>dpkg-source</prgn> - before
            applying the changes it will create the <file>debian</file>
            directory, and afterwards it will make
            <file>debian/rules</file> world-executable.
          </p>
        </sect1>
      </sect>
    </appendix>

    <appendix id="pkg-controlfields">
      <heading>Control files and their fields (from old Packaging Manual)</heading>

      <p>
        Many of the tools in the <prgn>dpkg</prgn> suite manipulate
        data in a common format, known as control files.  Binary and
        source packages have control data as do the <file>.changes</file>
        files which control the installation of uploaded files, and
        <prgn>dpkg</prgn>'s internal databases are in a similar
        format.
      </p>

      <sect>
        <heading>Syntax of control files</heading>

        <p>
          See <ref id="controlsyntax">.
        </p>

        <p>
          It is important to note that there are several fields which
          are optional as far as <prgn>dpkg</prgn> and the related
          tools are concerned, but which must appear in every Debian
          package, or whose omission may cause problems.
        </p>
      </sect>

      <sect>
        <heading>List of fields</heading>

        <p>
          See <ref id="controlfieldslist">.
        </p>

        <p>
          This section now contains only the fields that didn't belong
          to the Policy manual.
        </p>

        <sect1 id="pkg-f-Filename">
          <heading><tt>Filename</tt> and <tt>MSDOS-Filename</tt></heading>

          <p>
            These fields in <tt>Packages</tt> files give the
            filename(s) of (the parts of) a package in the
            distribution directories, relative to the root of the
            Debian hierarchy.  If the package has been split into
            several parts the parts are all listed in order, separated
            by spaces.
          </p>
        </sect1>

        <sect1 id="pkg-f-Size">
          <heading><tt>Size</tt> and <tt>MD5sum</tt></heading>

          <p>
            These fields in <file>Packages</file> files give the size (in
            bytes, expressed in decimal) and MD5 checksum of the
            file(s) which make(s) up a binary package in the
            distribution.  If the package is split into several parts
            the values for the parts are listed in order, separated by
            spaces.
          </p>
        </sect1>

        <sect1 id="pkg-f-Status">
          <heading><tt>Status</tt></heading>

          <p>
            This field in <prgn>dpkg</prgn>'s status file records
            whether the user wants a package installed, removed or
            left alone, whether it is broken (requiring
            re-installation) or not and what its current state on the
            system is.  Each of these pieces of information is a
            single word.
          </p>
        </sect1>

        <sect1 id="pkg-f-Config-Version">
          <heading><tt>Config-Version</tt></heading>

          <p>
            If a package is not installed or not configured, this
            field in <prgn>dpkg</prgn>'s status file records the last
            version of the package which was successfully
            configured.
          </p>
        </sect1>

        <sect1 id="pkg-f-Conffiles">
          <heading><tt>Conffiles</tt></heading>

          <p>
            This field in <prgn>dpkg</prgn>'s status file contains
            information about the automatically-managed configuration
            files held by a package.  This field should <em>not</em>
            appear anywhere in a package!
          </p>
        </sect1>

        <sect1>
          <heading>Obsolete fields</heading>

          <p>
            These are still recognized by <prgn>dpkg</prgn> but should
            not appear anywhere any more.

            <taglist compact="compact">

              <tag><tt>Revision</tt></tag>
              <tag><tt>Package-Revision</tt></tag>
              <tag><tt>Package_Revision</tt></tag>
              <item>
                  The Debian revision part of the package version was
                  at one point in a separate control field.  This
                  field went through several names.
              </item>

              <tag><tt>Recommended</tt></tag>
              <item>Old name for <tt>Recommends</tt>.</item>

              <tag><tt>Optional</tt></tag>
              <item>Old name for <tt>Suggests</tt>.</item>

              <tag><tt>Class</tt></tag>
              <item>Old name for <tt>Priority</tt>.</item>

            </taglist>
          </p>
        </sect1>
      </sect>

    </appendix>

    <appendix id="pkg-conffiles">
      <heading>Configuration file handling (from old Packaging Manual)</heading>

      <p>
        <prgn>dpkg</prgn> can do a certain amount of automatic
        handling of package configuration files.
      </p>

      <p>
        Whether this mechanism is appropriate depends on a number of
        factors, but basically there are two approaches to any
        particular configuration file.
      </p>

      <p>
        The easy method is to ship a best-effort configuration in the
        package, and use <prgn>dpkg</prgn>'s conffile mechanism to
        handle updates.  If the user is unlikely to want to edit the
        file, but you need them to be able to without losing their
        changes, and a new package with a changed version of the file
        is only released infrequently, this is a good approach.
      </p>

      <p>
        The hard method is to build the configuration file from
        scratch in the <prgn>postinst</prgn> script, and to take the
        responsibility for fixing any mistakes made in earlier
        versions of the package automatically.  This will be
        appropriate if the file is likely to need to be different on
        each system.
      </p>

      <sect><heading>Automatic handling of configuration files by
      <prgn>dpkg</prgn>
        </heading>

        <p>
          A package may contain a control information file called
          <tt>conffiles</tt>.  This file should be a list of filenames
          of configuration files needing automatic handling, separated
          by newlines.  The filenames should be absolute pathnames,
          and the files referred to should actually exist in the
          package.
        </p>

        <p>
          When a package is upgraded <prgn>dpkg</prgn> will process
          the configuration files during the configuration stage,
          shortly before it runs the package's <prgn>postinst</prgn>
          script,
        </p>

        <p>
          For each file it checks to see whether the version of the
          file included in the package is the same as the one that was
          included in the last version of the package (the one that is
          being upgraded from); it also compares the version currently
          installed on the system with the one shipped with the last
          version.
        </p>

        <p>
          If neither the user nor the package maintainer has changed
          the file, it is left alone.  If one or the other has changed
          their version, then the changed version is preferred - i.e.,
          if the user edits their file, but the package maintainer
          doesn't ship a different version, the user's changes will
          stay, silently, but if the maintainer ships a new version
          and the user hasn't edited it the new version will be
          installed (with an informative message).  If both have
          changed their version the user is prompted about the problem
          and must resolve the differences themselves.
        </p>

        <p>
          The comparisons are done by calculating the MD5 message
          digests of the files, and storing the MD5 of the file as it
          was included in the most recent version of the package.
        </p>

        <p>
          When a package is installed for the first time
          <prgn>dpkg</prgn> will install the file that comes with it,
          unless that would mean overwriting a file already on the
          file system.
        </p>

        <p>
          However, note that <prgn>dpkg</prgn> will <em>not</em>
          replace a conffile that was removed by the user (or by a
          script).  This is necessary because with some programs a
          missing file produces an effect hard or impossible to
          achieve in another way, so that a missing file needs to be
          kept that way if the user did it.
        </p>

        <p>
          Note that a package should <em>not</em> modify a
          <prgn>dpkg</prgn>-handled conffile in its maintainer
          scripts.  Doing this will lead to <prgn>dpkg</prgn> giving
          the user confusing and possibly dangerous options for
          conffile update when the package is upgraded.</p>
      </sect>

      <sect><heading>Fully-featured maintainer script configuration
      handling
        </heading>

        <p>
          For files which contain site-specific information such as
          the hostname and networking details and so forth, it is
          better to create the file in the package's
          <prgn>postinst</prgn> script.
        </p>

        <p>
          This will typically involve examining the state of the rest
          of the system to determine values and other information, and
          may involve prompting the user for some information which
          can't be obtained some other way.
        </p>

        <p>
          When using this method there are a couple of important
          issues which should be considered:
        </p>

        <p>
          If you discover a bug in the program which generates the
          configuration file, or if the format of the file changes
          from one version to the next, you will have to arrange for
          the postinst script to do something sensible - usually this
          will mean editing the installed configuration file to remove
          the problem or change the syntax.  You will have to do this
          very carefully, since the user may have changed the file,
          perhaps to fix the very problem that your script is trying
          to deal with - you will have to detect these situations and
          deal with them correctly.
        </p>

        <p>
          If you do go down this route it's probably a good idea to
          make the program that generates the configuration file(s) a
          separate program in <file>/usr/sbin</file>, by convention called
          <file><var>package</var>config</file> and then run that if
          appropriate from the post-installation script.  The
          <tt><var>package</var>config</tt> program should not
          unquestioningly overwrite an existing configuration - if its
          mode of operation is geared towards setting up a package for
          the first time (rather than any arbitrary reconfiguration
          later) you should have it check whether the configuration
          already exists, and require a <tt>--force</tt> flag to
          overwrite it.</p></sect>
    </appendix>

    <appendix id="pkg-alternatives"><heading>Alternative versions of
        an interface - <prgn>update-alternatives</prgn> (from old
    Packaging Manual)
      </heading>

      <p>
        When several packages all provide different versions of the
        same program or file it is useful to have the system select a
        default, but to allow the system administrator to change it
        and have their decisions respected.
      </p>

      <p>
        For example, there are several versions of the <prgn>vi</prgn>
        editor, and there is no reason to prevent all of them from
        being installed at once, each under their own name
        (<prgn>nvi</prgn>, <prgn>vim</prgn> or whatever).
        Nevertheless it is desirable to have the name <tt>vi</tt>
        refer to something, at least by default.
      </p>

      <p>
        If all the packages involved cooperate, this can be done with
        <prgn>update-alternatives</prgn>.
      </p>

      <p>
        Each package provides its own version under its own name, and
        calls <prgn>update-alternatives</prgn> in its postinst to
        register its version (and again in its prerm to deregister
        it).
      </p>

      <p>
        See the man page <manref name="update-alternatives"
        section="8"> for details.
      </p>

      <p>
        If <prgn>update-alternatives</prgn> does not seem appropriate
        you may wish to consider using diversions instead.</p>
    </appendix>

    <appendix id="pkg-diversions"><heading>Diversions - overriding a
    package's version of a file (from old Packaging Manual)
      </heading>

      <p>
        It is possible to have <prgn>dpkg</prgn> not overwrite a file
        when it reinstalls the package it belongs to, and to have it
        put the file from the package somewhere else instead.
      </p>

      <p>
        This can be used locally to override a package's version of a
        file, or by one package to override another's version (or
        provide a wrapper for it).
      </p>

      <p>
        Before deciding to use a diversion, read <ref
        id="pkg-alternatives"> to see if you really want a diversion
        rather than several alternative versions of a program.
      </p>

      <p>
        There is a diversion list, which is read by <prgn>dpkg</prgn>,
        and updated by a special program <prgn>dpkg-divert</prgn>.
        Please see <manref name="dpkg-divert" section="8"> for full
        details of its operation.
      </p>

      <p>
        When a package wishes to divert a file from another, it should
        call <prgn>dpkg-divert</prgn> in its preinst to add the
        diversion and rename the existing file.  For example,
        supposing that a <prgn>smailwrapper</prgn> package wishes to
        install a wrapper around <file>/usr/sbin/smail</file>:
        <example>
   dpkg-divert --package smailwrapper --add --rename \
      --divert /usr/sbin/smail.real /usr/sbin/smail
        </example> The <tt>--package smailwrapper</tt> ensures that
        <prgn>smailwrapper</prgn>'s copy of <file>/usr/sbin/smail</file>
        can bypass the diversion and get installed as the true version.
        It's safe to add the diversion unconditionally on upgrades since
        it will be left unchanged if it already exists, but
        <prgn>dpkg-divert</prgn> will display a message.  To suppress that
        message, make the command conditional on the version from which
        the package is being upgraded:
        <example>
   if [ upgrade != "$1" ] || dpkg --compare-versions "$2" lt 1.0-2; then
      dpkg-divert --package smailwrapper --add --rename \
         --divert /usr/sbin/smail.real /usr/sbin/smail
   fi
        </example> where <tt>1.0-2</tt> is the version at which the
        diversion was first added to the package.  Running the command
        during abort-upgrade is pointless but harmless.
      </p>

      <p>
        The postrm has to do the reverse:
        <example>
  if [ remove = "$1" -o abort-install = "$1" -o disappear = "$1" ]; then
     dpkg-divert --package smailwrapper --remove --rename \
        --divert /usr/sbin/smail.real /usr/sbin/smail
  fi
        </example> If the diversion was added at a particular version, the
        postrm should also handle the failure case of upgrading from an
        older version (unless the older version is so old that direct
        upgrades are no longer supported):
        <example>
  if [ abort-upgrade = "$1" ] && dpkg --compare-versions "$2" lt 1.0-2; then
     dpkg-divert --package smailwrapper --remove --rename \
        --divert /usr/sbin/smail.real /usr/sbin/smail
  fi
        </example> where <tt>1.0-2</tt> is the version at which the
        diversion was first added to the package.  The postrm should not
        remove the diversion on upgrades both because there's no reason to
        remove the diversion only to immediately re-add it and since the
        postrm of the old package is run after unpacking so the removal of
        the diversion will fail.
      </p>

      <p>
        Do not attempt to divert a file which is vitally important for
        the system's operation - when using <prgn>dpkg-divert</prgn>
        there is a time, after it has been diverted but before
        <prgn>dpkg</prgn> has installed the new version, when the file
        does not exist.</p>

      <p>
        Do not attempt to divert a conffile, as <prgn>dpkg</prgn> does not
        handle it well.
      </p>
    </appendix>

  </book>
</debiandoc>
<!-- Local variables: -->
<!-- indent-tabs-mode: t -->
<!-- End: -->
<!-- vim:set ai sts=2 sw=2 tw=76: -->