This GEM-free UMS code-path was never merged though into the xf86-video-intel DDX, as it would add about 50,000 lines of code into this open-source X.Org driver and would likely receive little in the way of work and testing. Adding back this UMS code-path also didn't solve all of the problems nor does it address the fundamental issue of KMS/GEM not working well for the old Intel chipsets.
Last month there was a new proposal by Intel's Chris Wilson adds ShadowFB to Intel kernel mode-setting so that you can benefit from KMS with clean mode-setting, a nice display experience, etc. But with using a shadow frame-buffer you will lose any form of 2D/3D/video acceleration. Adding ShadowFB support to the Intel DDX driver though is much less invasive than adding back an entire UMS code-path to the driver so the support burden is much less and it is another bypass to the i8xx cache coherency issues.
While Canonical has been interested in these measures to better the i8xx IGP support in Ubuntu, none of these new changes will work their way into Ubuntu 10.10. There's just a month left to the release of Ubuntu 10.10 "Maverick Meerkat" and we're already into a feature freeze, which is leaving developers scurrying to decide what to do about the Intel hardware support.
Christopher James Halse Rogers is now proposing that Ubuntu either blacklists KMS on the problematic IGPs and thus let the VESA driver handle those i8xx users, drop the i8xx hardware from the X auto-load list to let fbdev take over, or do nothing and let the Intel driver continue functioning with KMS along its normal code-paths. The discussion surrounding this for Ubuntu is on their X mailing list.
The option being most considered at this point is kicking Ubuntu users with Intel i8xx hardware back to using the VESA driver. Of course, this means no form of GPU acceleration for anything, poor user-space mode-setting, and no real display features at all.