Chris points out in his blog entry that Xgl is a lot of work:
[...] The big thing about the AIGLX approach is that it's incremental. The path to get from here to there allows us to leverage the strengths we have, allow people to participate as their hardware allows and work iteratively with vendors to add support to cards. Simply put, AIGLX presents us with a hill instead of a cliff. That's the big difference.
But it is not an academic discussion anymore about the complexity of getting from here to there. We are already there, and the code is already on CVS and we have already climbed the most difficult part of the cliff.
Anyways, Chrisl, what about addressing the comments on the Fedora Wiki to reflect David's comments? There is a link added, but there were no "architectural" changes developed in private.
Anyways, more power to the Red Hat folks with AIGLX, am sure we will have a combination of both in various machines and am just happy to be able to use both depending on the hardware I have available at my disposal.