Graesslin Has Compositing Dreams, But is it Yours?

by Susan Linton - Apr. 04, 2011Comments (3)

Martin Graesslin has a dream. On the anniversary of the great Martin Luther King, Jr.'s death, it's appropriate to note its importance pales in comparison. Graesslin would like to devise a way to have an "always composited desktop." As far as open source dreams go, it's admirable, but is it advisable?

There are times when one might want to disable compositing. Graesslin's example is in the case of saving battery life. Other's might be when starting a 3D game or other heavy applications, watching movies, or with older or lower resource machines. For these examples, KWin's usual method of unredirecting, or disabling composite on a per application basis while the actual effects engine is still running in the background, might be ineffective, counter-productive, or unsupported in a given application. While Alt+Shift+F12 can turn it off, most users don't know of it or want the hassle of knowing when to use it. So, Graesslin thinks something else should be done.

Coincidently, a colleague has had the same thought and already begun the work. So far under the new method, applications can suspend compositing as whole, stopping the engine in the background as well. This will reduce artifacting and tearing while improving performance. This sounds all well and good, but continuing the trend of desktops across the board these days, user control is being removed and handled automatically under the hood.

Removing the user control in System Settings is a mistake as no provision or consideration is given for slower hardware or uncommon setups such as dual monitors. Not everyone runs applications in fullscreen either. And most significantly, not everyone wants to run compositing on their desktop and they should be able to turn it off (or on) if they choose. In addition, applications will have to be patched to support this new feature and not all will, especially older games that no longer receive updates. The new system might be an improvement in some significant ways, but removing any user control will cause end-users problems and leave KDE open to lots of unintended consequences.

Developers seem little concerned with end-users that fall outside of their target demographic though. Graesslin's co-developer said in the comments to someone asking about older hardware, "Of course KWin will only allow compositing on systems which support it. But it is not our fault if users want to run the latest software on old hardware. Some day we have to start thinking about not supporting legacy hardware." Fortunately, he did go further and say that the Alt+Shift+F12 option would remain for those that know of it, but automatically handling features while removing user choice and control is an ugly path and a troubling trend to pursue.

Graesslin hopes to have all this integrated by KDE 4.8.

John Mark Walker uses OStatic to support Open Source, ask and answer questions and stay informed. What about you?


Failed automation is the downfall of Windows and most out of box distros. I don't see any reason why it shouldn't be possible to have a default use case as above but still provide a full over-ride in settings for power users.

I hate the thought of Linux going down then "we know better than the end user" path. Next there'll be a paperclip in Libreoffice " looks like you're writing a letter.." Pah.... DON'T DO IT!

This suggestion. Ubunty with unity, xorg replacements with less functionality than originally designed... It's not looking good.

It makes sense for commercial entities to restrict and remove functionality and support the 80% demographic, but that isn't something FOSS needs to worry about and there is no valid reason to start providing anything less than 100%.

0 Votes

I do not see any downfall in Microsoft. I am not fan of them but what i see is that there are market per everyone.

0 Votes


I think Martin's intro got many ppl. into the wrong mood.

a) KWin has a mighty rule system (hopefully a little more enduser compatible in the next release ;-) and WILL allow you to downforce this behaviour in any direction. (to be more precise: it does in git master)

b) "Of course KWin will only allow compositing on systems which support it. But it is not our fault if users want to run ..." quotes Martin, not me. (and he's btw, right, see below)

c) The UI removal of the "suspend" button is no big deal at all - it's hard to find anyway and might easily be replaced by some plasmoid or whatever since the dbus hooks remain.

d) personally i don't think we'll remove the option to start kwin w/o compositing active too soon or ever on X11 times.

e) even iff (and given the present and long time GL driver quality i would object such move) the option was removed, the rule systems would still provide a way to completely disable compositing (by setting a blind blocker that applies to all windows)

Goal of this blocking system is not to take the user out of control but to allow clients to say "hey, i'm really resource intensive and while i'm active there won't be any use for compositing anyway - so you could just shut it down and allow me to paint some more fps)

If a client would abuse this setting i'd first report a bug there and then simply get rid of it (or - if it's reeaallly important:: add a rule to fix it ;-)

Nevertheless, Martin is terribly right in one point that everybody should start to get past:

Currently X11 is the display server the least fitting average user demands.

Yes there is network transparency (which is completely overestimated: todays client/server software operate on data sockets or tcp - only real legacy software requires the X11 network to share licenses, and otherwise there's hardly any reason to run a client over a network on today *nix which provide shared libs and use tons of pixmaps in Gtk+/Qt & frineds) but the average user will easily kick it for window management à la exposé.

So there /will/ be some replacement and this replacement (which you don't have to use) /will/ have integrated compositing (because the way it's done now is a great achievement of Keith on the limitations of a X11 extension, but completely silly as well) what mens that you will NOT be able to use it if your GPU has just enough memory to keep one screensize frame.

Like it or not, but that's the way it is ;-P

0 Votes
Share Your Comments

If you are a member, to have your comment attributed to you. If you are not yet a member, Join OStatic and help the Open Source community by sharing your thoughts, answering user questions and providing reviews and alternatives for projects.

Promote Open Source Knowledge by sharing your thoughts, listing Alternatives and Answering Questions!