How Important Is The Wayland Display Server?

[pabloski]

pabloski: Gallium3D is a framework for graphic drivers. It's only purpose it to draw stuff (no matter if it's lines, textures, triangles, video or whatever).
Xorg has other components, like input devices, the X protocol, window management (with the help of a WM) and a lot of glue in between.


G3D is important as it allows many features the classic mesa architecture didn't permit, but as bridgman said, it's not meant to replace anything but the gfx drivers.

I agree, but currently with have a lot of apis (xrender, mesa, xvideo) that tries to compete to access the graphics hardware

this competition causes the problems with all know ( videos that doesn't work as expected under compiz for example )

gallium controls the hardware and abstracts it, forcing all the apis to go through the same "states" mechanism

and also gallium and kms will strip all the dangerous hardware voodoo tasks from xorg

I think this will improve X a lot

[bridgman]

You ignored something mentioned in this thread, namely that X is too complex for many people to understand and therefore to fix and patch.

I keep hearing that, however the most complex and problematic parts are going to be re-used by Wayland. Remember that krh is one of the developers who has been working on the driver stack used by both X and Wayland.

A lot of the "complex" examples cited in the Google Groups thread (DRI, VA-API etc..) are not part of X, however they *will* probably be used with Wayland.

I think Wayland is a good idea, and I'm not opposed to criticism of X; I just think a lot of the criticism is missing the point. It's fun to blame the older, more stable parts of X, complain that nobody is working on X, and conclude that replacing X is the way to fix stability problems, but the reality is that the most problematic parts are being reused by Wayland. There is also a *lot* of work going into modernizing and cleaning up the problematic parts (primarily input and output drivers) which will benefit both X and Wayland.

[bridgman]

I agree, but currently with have a lot of apis (xrender, mesa, xvideo) that tries to compete to access the graphics hardware

this competition causes the problems with all know ( videos that doesn't work as expected under compiz for example )

gallium controls the hardware and abstracts it, forcing all the apis to go through the same "states" mechanism

and also gallium and kms will strip all the dangerous hardware voodoo tasks from xorg

I think this will improve X a lot

It's really the modesetting code which requires X to have root access and which results in competing access to the hardware, not acceleration. The acceleration code in our open source X drivers doesn't touch hardware at all on 6xx and up, and only touches hardware on older GPUs when the drm is not available. Competing requests between acceleration clients have been handled by the drm for a long time, and the same mechanism is being used in Wayland.

I don't really think Gallium3D is going to do much to make X more robust -- it just replaces the simple, and fairly stable driver code in X with calls into the more complex and problematic 3D stack. It's still worth doing but don't read too much into it. I still believe KMS will do more to stabilize X than Gallium3D.

Gallium3D on its own won't do anything for video over Compiz either. The problem there is not multiple APIs or multiple clients, it's that compositors have not been "designed into" the display stack but rather plug in as optional components. As a result, we don't have synchronization mechanisms which work properly across a compositor - although that is one of the things the driver devs are working on.

Anyways, I guess a good summary might be :

Everyone agrees there are problems with the current graphics stack, but the most problematic areas (a) are not part of X, (b) are going to be re-used by Wayland, and (c) are being aggressively fixed, which will benefit both X and Wayland. There is a lot of old crufty code in X but it generally works, which is why nobody is worrying about it too much.

A number of new APIs have been added (eg XRandR) which could allow the removal of older APIs once the new ones are fully implemented; it's the dropping of legacy APIs and legacy hardware support which allows simplification of the code, not "fixing an old broken design". So far the X user community has not been open to the dropping of legacy APIs and hardware support, which may end up forcing a fresh start either in the form of something like Wayland or as a new version of X which drops all the legacy support and lives alongside the current code base in a much simplified form.

[bugmenot]

Nice to see some quality discussion on this topic. I am sure that the FOSS community/leaders will choose the right path in terms of Xorg/Wayland.

[curaga]

Dropping working hardware support should never be done IMHO. It doesn't even take space on modular installations, removing that argument.

Isn't one of the strong points of linux that devices old enough to be unsupported elsewhere still work here?

[89c51]

even if all the "problems" of X are related to the underlying infrastructure (i know that people work on it) IMO the code simplification and "modernization" -which is the main purpose of wayland- will only be benefitial to the FOOS desktop



the only thing that it is not included in wayland is network support

[89c51]

Nice to see some quality discussion on this topic. I am sure that the FOSS community/leaders will choose the right path in terms of Xorg/Wayland.

only thinking the situation with package management having more options in the Linux community can only be scary

[nanonyme]

Or those of us that have the displeasure of messing around with xserver/xorg.conf. I've had to deal with modelines and EDID options and resolution switching not working and absurd amounts of crap with it over the years.

Heh, thanks to opensource development, I haven't had to even use a xorg.conf for half a year or so.
Edit: The good multiscreen support apparently requires KMS. :3

[energyman]

'clutter backend'? why not add gvfs and pulseaudio crap too!

seriously, why do some devs think it is ok to infest EVERYTHING with gnome?

[nanonyme]

I think Wayland is a good idea, and I'm not opposed to criticism of X; I just think a lot of the criticism is missing the point.

I still stand by my view that Wayland is a bunch of good ideas but Wayland as a replacement probably is a very bad idea. We'd lose more than we'd gain if we'd replace the X server with something that has reduced functionality by design. (that said and having read MAD's post about X12, I think Wayland might play an important part in figuring out what X12 should be like)