I removed the GTK story from osnews today after privately discussing the issue with the people involved. But there is a bet that I will keep.
Owen Taylor promised that the final version of GTK will be optimized and I promised back that if it is, I will give them 50,000 free ads on osnews for gnome or fdo. But if it is more than 15-20% slower than gtk 2.6, I will put a new story back on osnews and this time I will be even more strict in the issue.
I take performance and stability very seriously and I don't take shit for it. GTK's current performance is already poor compared to Qt's (responsiveness sucks), so adding the Cairo stuff on top is the last thing I want as a user. I AM willing to sacrifice 15-20% of the performance for the new goodies Cairo brings, but more than that is just unacceptable at this point in time.
No, GLitz/XGL won't be ready for Gnome 2.12 in September, so people should stop using that argument. If GTK 2.8 ships with Cairo and gnome 2.12, it will be using software emulation and performance will suck immensely if Owen won't manage to optimize the shit out of it. Nautilus already depends on Cairo for example.
That's where my problem lies: I don't believe it can/will be optimized as much as needed. Call me a pessimist, but I will believe it when I see it. Owen is a good guy --I am sure--, but he doesn't strike me like the guy who will optimize things. He hasn't done so for GTK all these years, I don't believe that he will start doing so now.