I mailed our faculty mailing list touting my textbooks (Diffy Qs and Real Analysis), and one of the responses linking this post gave me an idea. Perhaps I should start a sort of sourceforge.net for math textbooks. Simply set up a repository (svn would work best I assume). Then set up some web around it. The site would definitely need some “editorial board” that would approve what textbooks are allowed, and there should be some sort of way to make versions that are “blessed” by the board and that would then be in some sense “stable.” Books could be in development and of course branches could be set up to work towards the next “blessed” version of the book.
I would definitely want to avoid the problem of wikipedia, wikibooks, planetmath, and such, where anyone off the street can come in and start typing what they think is math. The problem of wikipedia is of course that anyone can do whatever they want. Planetmath was touted as being more strict and having maintainers for topics. The problem with planetmath is that the way maintainership is decided is on a first come first serve basis. If the homeless guy living in the local mall parking lot decides that he wants to write a topic on strictly pseudoconvex manifolds, then he will control that topic from now on. Regardless of qualification.