The BNF spec doesn't always agree with the draft and fails to put any structure onto the music elements - it roughly speaking defines a music line as a collection of things that can appear on a music line without imposing any order on them. Again by way of example, broken groups like A>B should be a syntactic unit, rather than <note> and <broken> just being something that can appear in a music line. (In part some of the problems stem from abc's mixing of line-based definitions - having to match s: and w: lines with the music lines - before you can deal with the music - you can get a better grammar if you preprocess them into the music).
If anyone's interested in going back and looking at the version 2 standard (or something in between 1.6 and 2) I'd be interested in getting involved. (I think the list died some years ago).
BTW, I don't want anything I've written above to imply that I don't like abc - I love the stuff and especially for single melodic lines it's hard to beat. I did the abc for the Yorkshire Garland project and for that I generated abc from musicxml extracted from a pdf produced by Sibelius - it's not as pretty as hand coded abc, but it was quick and accurate. And for storage of tunes it neat and compact - I implemented two versions of song browsers (one in Java, one in Prolog) for using the Digital Tradition database and stored the tunes there in abc (another conversion job I did from Songwright - I've suggested the DT changes to abc but they are reluctant!) . On the other hand, I could do without being able to arrange Beethoven in abc - I write and arrange a lot of classical guitar music and I do that in Finale or Sibelius, not abc. Just the ability to do it doesn't make it the right tool for the job!
Glad to see the site's up and hope it generates lots of interest.