you would think that... but as soon as you build applications against the nvidia headers.. you see such a rational thought contradicts reality.
Well that sucks. Sounds like a bug that should be fixed by Nvidia then.
How do you make use of the Nvidia extensions with the Mesa headers?
You dont.
if you really really really want to tie your binary to the existence of nvidia hardware and libs.. you use nvidia's headers and libs at compile time. And with livna's rpms you can do exactly that by setting appropriate -L and -I to point to the nvidia libs and headers when you are building/linking at compile time. This however is seldom something a community contributor wants to do.
I don't know what kind of extensions are offered. If some of them optimize for nvidia hardware then it makes perfect sense to me.
I think at this point... your questions are best answered by rebuilding sourcecode into binaries for yourself and watching the truth unfold.
Heh, okay I give up on this.