I tried using lilypond ( http://www.lilypond.org/ ) for typesetting of sheet music. While the output looks nice, the input language IMHO is quite horrible, because the underlying data/execution model is underspecified. For some parts, it tries to describe the logical structure of the score; but for others, the layout; and in addition it has several non-obvious context-dependencies (but see below), preventing modularity. Is there a better option? E.g., starting from a clear mathematical model, as in Haskore, and use lilypond only as a PDF rendering engine? Do I want hly / hts perhaps? http://rd.slavepianos.org/?t=hly As I see it, the main high-level design problem is that the source language needs partial evaluation annotations for abstractions applications: sometimes they should be expanded (for MIDI rendering, always) and sometimes not (in typesetting, to create repetition marks instead of actually repeating notes). PS: I agree that some of lilypond's context dependencies (relative pitch, implicit note length) do really save large amounts of tedious typing: "c4 e g a c1" is much more economical than "[c 1 qn, e 1 qn, g 1 qn , a 1 qn, c 2 fn]" which I guess is the Haskore equivalent.