--- /dev/null
+General format of a construct:
+
+ BLOCKNAME { <info to go with this block> }
+
+Some types allow declarations:
+
+ IDENTIFIER = BLOCKNAME {
+ <info to go with this block>
+ }
+
+ ..
+
+ BLOCKNAME {
+ IDENTIFIER
+ ...
+ }
+
+
+In musicmode, eg,
+
+ ''!c8.-"text"_v
+
+
+a lot of characters parse differently
+than in "command" mode, eg,
+
+ identifier = score { .. }
+
+So you have to signal that to the tokenizer. This is done with
+'$'. '$' is a delimiter, which used by the tokenizer only.
+
+* musicmode: The brace still is used to group grammatical groups.
+* musicmode: "word" are preceded by a '\' (backslash)
+
+This means you can write some stuff in a zillion ways:
+
+1. $\var = \blockname { ... } $
+
+2. var = blockname { $ ... $ }
+
+3. var = $ $ $\blockname { ... $ }
+
+COMMENTS
+
+not really crystallized; you can use '#' or '%' as line comment
+
+OTHER
+
+A correctly parsed .ly does not guarantuee output. A lot (most) of the
+checking is done *after* parsing (some checks are after the break calc!)
+I'm sorry.
+
+The parser's job is to construct appropriate objects. It will *only*
+detect parse errors.
+
+LilyPond first reads 'symbol.ini', which contains declarations crucial
+to proper operation of LilyPond (symbol tables)
+
+This language looks a lot like Rayce's which in turn owes a lot to the
+POVRay raytracer. Now, I know, musictypesetting and Raytracing do not
+necessarily require the same input format, but I was just to lazy to
+make up a new/better input format. Suggestions welcome.