Michael, I hope you're not shooting yourself in the foot about nVidia's conference. But you usually have been spot on the money in the past, so I'll have to wait and see what comes out regarding Linux from nVidia.
I was not expected that all X devs goes to Intel and now even Ian Romaniac is. He will (for sure), step-by-step, aligning Mesa to work on "every" platform with Larrabee, but not... Who cares for busted & delayed X experiment? Or for OGL3? Ian's & Transgaming's extensions? HAHAHAHA!!!
people are always disappointed. If the new release would have broken all the apis for the sake of 'clean up' a lot more people would be disappointed.
Whatever you do, you will always see some whining. Just look at this forum and the whining bridgeman& co have to endure (and endure they do, with humour)
No, in fact they have a good reason to be upset. A year ago a new api was promised, examples were shown in a newsletter and it would all be wonderfull: a modern, clean, easy to develop for api. An api that could be on par with DirectX 10. Now, a year later, after being quiet for almost a year (with no reasons given why OpenGL 3.0 was delayed) they come up with something that they should have called OpenGL 2.2.
Is it al that bad? Well... there are some positive points. There is a new depricated model that shows almost any fixed pipeline function to be depricated. This means, all depricated functionality being removed in the next version of OpenGL. It also means that it will take until the next version of OpenGL before it really gets easier for driver developers and application developers to develop OpenGL applications. Khronos members told in the opengl forums that the object oriented api is still being worked on, but the problem is that nobody trusts Khronos anymore after a year delay, a year of silence and a totally different api then promised.
Another thing that is different then promised is that you need DirectX 10 capable hardware for OpenGL 3.0 (initial plans were based on minimum R300/NV30). But probably the most important feature: geometry shaders still need an extension. So it is the question if Ati, Nvidia and Intel will support this extension.
Anyway: if you use the forward-compatible profile (profiles are also new), which doesn't use any depricated functions, you have a cleaned up api. But it would be nice if Khronos released a spec without all the depricated stuff in it (which takes up about 2/3 of the spec).
I read the reactions of a bunch of DX fanboys on opengl.org. The bottom line however, is that if they wish to sell their games to me, they better not use DX 10+ (should stick with OpenGL or DX 9), due to the Vista+ only support. Some game developers understand this perfectly well.
I think one problem with OpenGL is that it actually was never made for programming games with it. It was a tool to create powerful CAD applications, which didn't even need to be hardware accelerated. Because of its nature, OpenGL now keeps the support for the workstation business and tries to add support for games, but these two things actually have very different needs and thus, creating a "clean" API is very hard as there will always be areas which never get touched in either games or CAD programs.
Ever wondered why D3D is superior to OpenGL today? It's only supposed to be used for games, and as such the API is optimized for that use.
What we would need is a complete restart (yeah, I'm one of these guys who say starting from scratch is always a good idea ;-) ) with a new API that is tied to programming games only and let the OGL ARB do whatever they want to go back to workstation-only. The only problem with this approach is that it would probably damn hard to get the support from major companies like ATI or NVIDIA for that (however, that reason has never prevented some OSS developers to implement their ideas).