I don't know anything about GI, but given the small size of its community forum, I think it is pretty safe to assume that it has a very small user base and can do pretty much what it likes from one version to the next if the core group of users agrees to it.
JSF on the other hand has a very large user base with uncounted millions of lines of legacy code over the last decade and can't afford to do a complete redesign that will break backwards compatibility all over the place. Internal refactoring of the code is certainly possible, but if you start to make fundamental changes to the architecture that are exposed to external interfaces then you're usually better off starting (yet another) new framework entirely.
Skimming the GI 4.0 document, I don't really see anything extraordinary about it. It seems to be a mishmash of ideas from the main developer mixed in with a discussion of some of the requirements and requests that users have made. It goes into plenty of detail, that's true, but it reads more like one person's stream of consciousness then a definitive strategy document.
I've only recently come back to JSF, so I don't know, but do you think that the existing JSF requirements and design processes are lacking in some way?
[Message sent by forum member 'tacitust' (tacitus_at_yahoo.com)]
http://forums.java.net/jive/thread.jspa?messageID=390545