Well, actually, most compiz features were first show in WinHEC 2003. Novell copied the demo 3 years later with the misguided XGL implementation.

https://www.youtube.com/watch?v=X0idaN0MY1U

The funny thing about XGL is how ahead of its time it was. In some way, it still is. Building desktop OS on top of game engine is a trend starting to take hold. Newer Linux Mesa GLAMOR acceleration is just that. Qt6 will brings this to tons of embedded devices near you. Modern SOC are more efficient when pushing GUI into 3D space.

That’s not actually true. The majority job of a GUI is text rendering. I recommend reading some of Raph Levien’s posts on the matter. It’s really hard to get quality GPU-accelerated text rendering. Games don’t have this requirement because they draw very little text and it’s stylistic so all the high-quality text rendering requirements are bypassed. Moreover they also probably remove the need for arbitrary Unicode whereas GUIs can frequently not make such a simplification. It’ll be interesting to see what performance gains are realized keeping in mind from a battery perspective the GPU may be sufficiently more power hungry that the efficiency gains are insufficient to make it more power efficient.

> but work on Compiz started AFTER Aero was shown.

I'm not sure it's necessarily true they started development after Aero. They released it 8 months later. Did it only take 8 months for the first release?

The author also omits beryl, which was the real beginning of compiz.

Maybe my history is fuzzy, but I seem to recall beryl being a fork of compiz which eventually came back into the fold.

Actually. It was my history that was fuzzy, according to Wikipedia you’re right.