Oh, really? OK, let's have a look at the numbers YOU quoted, then:
Now, apart from the obvious fact that anybody with a first degree in finger counting could acknowledge, one could also argue that the answers given to 'Image quality' are more related to the video playback business than to 3D performance.
OpenGL performance 17.3 %
Video playback/acceleration 15.1 %
Licensing/open source 11.5 %
2D performance 10.8 %
Image quality 10.3 %
Syspend/hibernate 9.7 %
Stability 9.4 %
Ease of installation/maint. 8.1 %
Display-related features 4.8 %
Multi GPU rendering 2.3 %
In the same lines seen in some posts in this thread, I could say that efforts to improve 3D performance are a waste of time given the scarcity of linux games and the absolute no need for a compositing desktop. Of course, there would be an army of people including gamers, professional and academic users of 3D graphics software and linux enthusiast that wouldn't be very happy.
I don't know what's the hard part in understanding that decoding HD MPEG-2 puts quite a stress on the CPU, and that having the hardware we bought to do it would be, uhm, reasonable. The _technical_ questions related to the convenience of implementing this now or doing it within a unified API encompassing MPEG-4 standards are valid and solid. Arguments based on what one would personally like to see are not.
PS. On the other hand, you guys in the states may have some extra time to lobby for this:http://news.slashdot.org/article.pl?.../01/08/2052212