Do you think Xorg should be rewritten from scratch?

[Pickup]

Many things looks temporary in xorg: the memory manager, the 3d driver model, the 2d acceleration method, video playback...

Years are passing and things seem to point nowhere, everybody continue to say "hey, new and better acceleration, wow!" but after 2 mins people start complainting about their new ITA/DMA 4950+1 or PerFidia GeoStrenght 9900&1/2 or OutTel 995GMX performing as a SVGA from late '80s.

And manufacturers are complainting too: Nvidia replaces parts of X; Intel rewrites parts of X; ATI tries to work with the existing stuff and it took years to achieve acceptable performance.

Then, why don't rewrite X completely, if the actual one have no good solutions for the modern desktop needs?

[deanjo]

Many things looks temporary in xorg: the memory manager, the 3d driver model, the 2d acceleration method, video playback...

Years are passing and things seem to point nowhere, everybody continue to say "hey, new and better acceleration, wow!" but after 2 mins people start complainting about their new ITA/DMA 4950+1 or PerFidia GeoStrenght 9900&1/2 or OutTel 995GMX performing as a SVGA from late '80s.

And manufacturers are complainting too: Nvidia replaces parts of X; Intel rewrites parts of X; ATI tries to work with the existing stuff and it took years to achieve acceptable performance.

Then, why don't rewrite X completely, if the actual one have no good solutions for the modern desktop needs?

It's getting to the point where this wouldn't be a bad idea. X is old and tired, with the present solution we are seeing more and more where they paint themselves into a corner resulting the need for massive revamps all to often IMHO. Same could be said for sound in linux as well.

[Melcar]

They should. Sound as well. I just don't get why they keep "patching" an infrastructure that is obviously too old and has outlived its usefulness. Soon things will start looking like freaking Windows in the pre-Vista days.

[Gentooer]

Since X is modular, isn't that what they're basically doing now by rewriting it in parts? DRI2 (maybe)replacing DRI. UXA/EXA replacing AAX. Gallium3D replacing whatever was there before. GEM replacing TTM.

I don't think it would be very easy to just rewrite the whole thing at once.

[FunkyRider]

If this way can help them get rid of "hey! I've come up with a new acceleration mode, forget about what we did yesterday!" thingy, count me in.

[RealNC]

Applications are using it so you can't just throw it away and rewrite something new. Actually, someone did write something new. DirectFB. No one wants to use it.

[tsuru]

http://www.youtube.com/watch?v=-oFxhqYn-g0 at about 9 minutes in.

[erdizz]

I think that there is now potential for cooperation on deep refactoring of xorg. I say "refactoring" because keeping the API is likely a good idea. It is the one part of xorg that works quite well, provided that the whole *nix desktop thing depends heavily on it. But it is becoming increasingly evident that everything besides the API can be safely thrown away.

In fact, the refactoring (including the "throwing away" part) has been happening for quite a while now, but in a very disorganised fashion: EXA, Xgl, AIGLX, nVidia proprietary solution, new components from Intel, and now this UXA thing, etc. Overlapping efforts are already a norm, and still the cumulative effect of these efforts is not satisfactory, because things simply don't work in many cases.

The amount of dissatisfaction with the state of xorg makes starting a new project quite reasonable, in my opinion. What's needed is a well-formulated plan on what to do and a fair public discussion. That's the way to start an organised community effort.

[Ex-Cyber]

Many things looks temporary in xorg: the memory manager, the 3d driver model, the 2d acceleration method, video playback...

Years are passing and things seem to point nowhere, everybody continue to say "hey, new and better acceleration, wow!" but after 2 mins people start complainting about their new ITA/DMA 4950+1 or PerFidia GeoStrenght 9900&1/2 or OutTel 995GMX performing as a SVGA from late '80s.

Rewriting Xorg wouldn't fix this; it would make it worse. There would still be all the problems associated with the general difficulty of writing driver code, plus an unproven architecture that would probably not end up being stable.