He's an unpaid volunteer coder. He tested with what was available to him, and he added an optional effect.
However, bypassing KWin's internal GL checks might be required for it to work. See this post:
It's easy to point fingers at people who do nothing and say "at least report it".
But he does a lot of work. So what if maybe he doesn't really have that many time to mess with testing a lot of configs, hey that's where the community can help. It's not like kwin just materializes hours before KDE releases -- anyone can run a trunk build of kwin (regardless of using KDE at all) and test it out.
Of course it would be nice if he went to THAT extra trouble, but hey, if all of us here and commenting on that blog did that much for the free desktop, it would probably rock even more.
Honestly even I get a little tired of trying combos of kernel+X+mesa to get free drivers rocking. And when you find that perfect combo, dang you discover that kernel broke some other hardware you had.
So here's my 0.02 (euro) cents.
Visible bugs get reported (I've filed some myself.) What we can't see, we can't report, obviously.
Volunteer developers don't owe you or anyone else anything! He is not obligated to run a gazillion distributions with trunk builds of Xorg and Mesa.
No! It's the paid Xorg/Mesa developers duty to test their drivers with with the most common composite window managers (currently Compiz and KWin, in the future Mutter as well). The current messy situation wouldn't have happened if the well paid developers did their job in the first place.
And in case there are too few paid Xorg/Mesa developers, maybe some Linux distributors should put some money where it actually counts instead of spending it on a design team that makes purple backgrounds and moves window buttons to the left side....
Most of the Xorg/Mesa developers are volunteers as well. If the paid developers did all the testing you want instead of working on drivers, the problem would take care of itself because the GL 2 support and new hardware support being discussed here probably would not exist in the first place.
Somewhere around 5x the current staffing level would make a lot of things possible. In the meantime, more communication and making more options "user-selectable" rather than "automatic based on extensions and blacklists" would probably help.
What's the problem ? A gazillion is an indeterminate, large number and AFAIK there are an indeterminate, large number of distros to choose from :D
to hell with it still, i hate blur anyway.
Query the driver about whether it supports feature xyz and only proceed if it does.
If you don't have the hardware to test on, that's a pretty good strategy, unless, of course, the driver lies about what it actually supports. There's no excuse for a driver claiming capabilities it doesn't have.