On 12/12/05, Delirium delirium@hackish.org wrote:
I haven't been heavily involved in the open-music community, but I've attended a few academic computer-music conferences on and off, and the people I've met are somewhat ambivalent about Lilypond. Basically, the main complaint is that it's intended to be a music-*presentation* language (or music *typesetting*), not a music-*representation* language. This makes it hard to use for a lot of things, like automatic manipulation, transformation to ther formats, and so on. It also makes it relatively hard to edit, since you have to specify a lot of the nitty-gritty presentation details yourself.
Eh, hard to edit compared to what? There are two classes of music (vs sound) formats.
One is symbolic formats used for typesetting, analysis, and scores for humans. (Examples are lilypond, ABC, humdrum's format, commercial typesetting systems, etc). These convey the music at a fairly high symbolic level, and make it difficult to attack high resolution performance information. They generally make many forms of useful transformation fairly easy, and often provide hinting for logical grouping (ties) and other aids which either help perfomers working off a typeset copy or machines doing analysis understand the natural grouping of the music without hearing it.
The other is time series formats used for electronic music performance. (Examples, MIDI, Csound ORCs, PD's graphical score dohookey).
It usually trivial to convert from the symbolic to a kludgey low resolution performance type. The reverse is darn near impossible because the non-audiable logical grouping is usually lost.
Generally doing notation work in the symbolic type is pretty easy while in the second type you end up getting mired in the details. (0-127 pressure? I don't care I just want it louder than before!)
Of course, it's a poorly-funded area with only a relatively small body of interested people working on software, so there aren't many good replacements. There is one system called GUIDO, that was promising but still very early the last time I heard about it; it may have a release by now, but I don't know of it. One problem is that most people who are interested in this sort of functionality just give in and use Sibelius or Finale, both fairly full-featured but proprietary commercial programs---an "open source Finale" hasn't yet appeared.
There is also an MPEG working group trying to develop an open standard for music representation, but it works slowly. The main stumbling block, besides being a working group in the first place, is that music representation is a much harder problem than it seems at first---there are a *ton* of variations in how people have written scores over the years, many of them conveying substantive information that needs to be captured somehow.
Eh, What problem do you think Finale solves? It is pretty much as bad as lilypond from the perspective of storing a musical performance rather than a representation for typesetting. Have you heard the midi it produces? Yuck. Mr. Roboto.
The problem is largely that they are conflicting goals. For performance, we'd best want a flexible event time series format. For editing we want a highly symbolic format (like sheet music). Performance has the extra challenge in that we must somehow specify how to turn the notation into sound.
For an human compatible score, which works with all your notational wishes we could use PDFs.. but you lose editing and layout flexibility. There is lilypond, but it fails to capture all the interesting notations that have been out there. I expect it will be impossible to capture all notation in a format that still has a symbolic understanding of the music.
As far as performance goes, Lilypond will produce reasonable midi, but getting fine detail into the midi is impossible while keeping a readable score. As I said, conflicting goals.
As far as the conversion of notation into sound, thats what the MPEG4 structured audio group has tackled. Uses what is effectively midi for its score. It's totally useless unless you're in to the electroacoustic music scene.
Not that midi is an ideal performance representation.. lots of silly limits, problems of serialization, and poor resolution hinder that. Open sound control ( http://www.cnmat.berkeley.edu/OpenSoundControl/) fixes all that, but gets us even further away from something which can be converted to a symbolic representation sutiable for music.