Since there are already some projects going through in this area (e.g. pwgl or inscore), wouldn't it make sense to try to integrate with these, or try to help them, instead of reinventing the wheel?
Perhaps, but consider this:
The performer I am working with is a percussionist, and excellent at sight-reading music. However, he's not by any stretch of the imagination a programmer, and the idea of giving him command-line compilation issues to deal with, or complex connectivity between packages, would kill the project straight away. From me he needs to receive, via email, a PD patch that will just work. If other libraries are "wrapped" into PD i.e. externals are made and integrated into a future PD-extended, then these might provide some practical options for me to work with classical musicians who aren't programmers (and the majority of them are not). However, for the time being I am limited to that which can be rendered by the current PD-extended straight out of the (in)box, without any modifications to the computer it is running on.
That is why I'm building a system that uses just GEM and a truetype font, which can be made into a single package and distributed to the performer of my piece. If I had institutional support perhaps I could envisage something more complex to work, but I have been unlucky in that respect. I could either give up, or try to find a practical solution that works both for me and for a non-computer geek classically trained player. I choose the latter because I want to make music.
Best, Ed