LCA: Catching up with X.org [LWN.net]

LCA: Catching up with X.org

Posted Jan 23, 2009 16:56 UTC (Fri) by kev009 (subscriber, #43906)
Parent article: LCA: Catching up with X.org

I bet the XFree86 guys are feeling like bigger idiots by the day. This is what happens when you try and mess with free software and put off good developers. Quite nice poetic justice as XFree86 fades to nothingness.

I do, however, hope that these new features can be made to work on BSD and other UNIX platforms in time. It would be a shame for FreeBSD, quite a nice kernel/OS itself, to be in the position Linux has been in for this time simply because of popularity.

(Log in to post comments)

LCA: Catching up with X.org

Posted Jan 23, 2009 19:09 UTC (Fri) by drag (subscriber, #31333) [Link]

It should be possible to make these things work on BSD.

The DRI2 interface is what you need to have in order to communicate between the userspace Mesa-based drivers and in-kernel memory management stuff.

So if the BSD folks are able to modify their kernels to support the DRI2 interface and it's limited set of operations then compatibility should follow.

Probably the biggest problem is that I suppose the BSD folks rely on X drivers heavily for their GUI stuff. With the DRI2/GEM/KMS/Gallium3D stuff the need for X.org's DDX drivers are evaporating.

That is in the future the need for things like xorg-drv-video-intel (or whatever) will be gone. Instead of running your GUI as root you should be able to run it using normal everyday APIs that are available to any other application under your user account.

So unless the BSD folks get their butts into gear then they will get left behind and end up having to force to deal with a huge amount of Linux-isms... It's going to be very difficult for X.org to avoid shovelling to much Linux-specific code when the interfaces they need in BSD are simply non-existent.

-----------------------------------------

One of big features for Vista, for example, was a new userspace driver model for graphics. As part of it they touted a new kernel memory manager for graphics. This was one of the big issues that early Vista users had... their video drivers from XP wouldn't work that well and the new vista-specific drivers were not mature.

And I expect that if anybody here was to look at OS X they would have a similar setup for their composited interface.

So I think that if you want to provide modern graphics and modern APIs that take advantage of GPUs this is just the sort of thing you need.

Plus moving your GUI out of being setuid root binary is always going to be a good thing.

LCA: Catching up with X.org

Posted Jan 23, 2009 21:15 UTC (Fri) by jamesh (subscriber, #1159) [Link]

Note that the commands being stuffed in the batch buffers are still very GPU-dependent. There is still a need for GPU-specific drivers with all this work.

LCA: Catching up with X.org

Posted Jan 24, 2009 0:11 UTC (Sat) by drag (subscriber, #31333) [Link]

Well that is what Gallium is designed to deal with right? Provide a winsys driver that is hardware dependent and then use that as a way to translate GPU-independent APIs to GPU-specific instructions...

Gallium3D should be introduced in a testable form in Mesa 7.5.

LCA: Catching up with X.org

Posted Jan 23, 2009 19:13 UTC (Fri) by branden (guest, #7029) [Link]

Having dealt, some years ago, on a regular basis with some of the "core"* XFree86 members, I am confident that they feel thoroughly vindicated by the course of events.

Yes, that's right. Best not to think about it too hard. ;-)

* XFree86 disbanded its core team contemporaneously with the mass exodus to X.Org; my guess is that they twigged correctly that there was no point having a "core" development team when there was not going to be a "periphery".

LCA: Catching up with X.org

Posted Jan 24, 2009 8:45 UTC (Sat) by flewellyn (subscriber, #5047) [Link]

So they deliberately sabotaged the project in order to spur someone else to do the improvements they didn't want to do?

That's not making sense to me.

LCA: Catching up with X.org

Posted Jan 26, 2009 13:29 UTC (Mon) by branden (guest, #7029) [Link]

No. I'll stop being so opaque.

I can't find the mailing list reference--and for all I know the message isn't archived anymore, but at one point one of the erstwhile core team members bluntly said (generously paraphrased) that the project had far too many users, was too Linux-centric, and that times were much better when they were just a small group of friends hacking up a descendant of X386 for their own fun and edification.

Now, I don't know if this opinion was shared by all of the core team, or even that this wasn't necessarily a sarcastic rant, but it does render XFree86's philosophy of project management comprehensible, and their actions rational within that framework.

The only wart on it is that these goals were not (and are not) stated forthrightly. XFree86's conception of its userbase was a mass of essentially anonymous people who could use the code if they wanted to, were welcome to file bugs or submit patches, but should not ever expect more than the most perfunctory conversation, or to ever enjoy elevation into the clubhouse. (That said, one developer, after many years, did achieve such an ascension, apparently through years of hard work and the forging of a personal friendship with a core team member, in part through trade shows).

XFree86 did deviate from that founding principle, leading to such things as (gulp) a public bug tracker, public betas, a second tier of half a dozen or so serious contributors, and even the injection of an X architect into the core team, but all of these served to poison the original vision.

By these lights, the X-Oz license, later the XFree86 1.1 license, makes perfect sense. By crafting a license whose terms were objectionable to most of the community and its developers (especially the smelly proletariat comfortable with the GPL and impure variants of the original BSD licenses which were missing clauses), but to which the trait of "freedom" could still arguably be said to apply--and thus positive relations with the beloved *BSD projects preserved--it succeeded in metaphorically nailing a "KEEP OUT" sign to the clubhouse door.

That even the *BSDs gave up on XFree86 is, under this model, regrettable but by no means an indication of failure. In a way, it's better this way, because the proceedings of XFree86's august society are even less frequently disturbed.