Today I read a news post about the KDE 3.5 fork Trinity and it was discussed that you should use it because of the bad performance of KWin:
the one constant is abysmal Kwin performance.
So to me it sounds like the author of said news article things it is a better idea to run the KDE 3.5 fork, because there KWin is without (OpenGL) compositing, instead of just turning off compositing in KDE 4. Well that’s interesting reasoning. (Btw. if you do not run at least 4.7.2 you should not complain about KWin performance).
Now to my actual blog post: I appreciate the idea of the Trinity developers to bring back the KDE 3.5 desktop experience to those users who really want it. This is a great offer. Personally I doubt that the Trinity project is doing the effort in the right approach. Instead of just porting Kicker and KDesktop to Qt4/KDE4 they forked everything and the kitchen sink. I rather doubt that a team which is smaller than the KWin team is able to maintain not only KWin but also every other part of KDE 3.5 and now also Qt 3.
The developers of Trinity offer the users something that their system cannot provide: a future for complete KDE 3.5. I did a lengthy analysis of Trinity for German readers in the freiesMagazin some time ago: Trinity – Desktop ohne Zukunft