Linked by Oscar Boykin on Wed 25th Feb 2004 08:59 UTC
X11, Window Managers For the past year, there has been a slow and steady stream of news events regarding XFree86, X11, or new X server implementations. To those not paying close attention (and even those who are), the meaning of some of these events may not be clear. In this brief article, I attempt to share my impression on what the changes mean for users of free software on the desktop. It appears that XFree86 is in some turmoil, and it may leave some to infer that free desktop systems will suffer.
Order by: Score:
by Uruloki on Wed 25th Feb 2004 09:18 UTC

Keith mentioned a few interesting things on FOSDEM last weekend. Looks like they'll be releasing new xfree86 code under the old (non-defunct) license in the near future.
Well, most of their code doesn't work at all, but still .. they're getting places.

A few comments...
by bytes256 on Wed 25th Feb 2004 09:35 UTC

Excellent article overall...I have a few comments...

First of all I never realized that XFree86 maintained all of their dependencies internally! No wonder it takes so long to compile. (I usually just start it and walk away whenever I have to compile it.)

such as converting the build system from imake to autotools

imake is definitely aweful, but I don't think autotools is really the best solution either. I would rather see them using something like scons - it is a far more productive build system and makes porting very easy. A project as large and complicated as XFree86 would be very difficult to switch over to autotools. Certainly not impossible, but difficult. Anyway, why not use the best technology available if you are going to go through the effort of something as significant as rewriting the build system from scratch.

I also agree that XFree86 needs to be a lot more modular. Most importantly, for god sakes open up the development process more, it shouldn't be harder to add code to XFree86 than it is to add code to Mac OS X! (I'm referring to Darwin.)

Execellent article !
by Eitan Kovacs on Wed 25th Feb 2004 09:41 UTC

this is an execellent article.
i think that the leadership of a new project should come from
the main linux distributions and the community.

by Anonymous on Wed 25th Feb 2004 10:08 UTC

XOuvert is not a fork but consider themselves as branch (just visit their webpage). And yes, together with will fork XFree86 starting with 4.4 RC2.

very good article
by retrakker on Wed 25th Feb 2004 10:16 UTC

i can just second the previous posts: a really good article - thats what i want to read on osnews. thanks

As usual ignorance leads the article.
by Anonymous on Wed 25th Feb 2004 10:16 UTC

Just asking ? Does it have a reason that you left out KDE here when you wrote the first paragraph ? .... like GNOME .... and where is KDE ? Not to mention that the guy who works hard on x f.d.o (Keith Packard) is a KDE supporter. Ok these are little details, but you should think of those who never heard about either KDE/GNOME or XFree86 before, just coming here reading this article.

By the way regardles of Keith Packard being thrown out (like many others). There is no demand being in a team or something specially when not getting paid for the work or not even being under contract. You as always have the chance to fork a project and continue it on your own with others in the hope that it becomes mainstram one day. That's these attepts like Xouvert, X Freedesktop, Y Windowing System.

Another point is, that I've been following the XFree86 development quite closely the past 1-2 years now and I must admit that they are doing pretty fine. The entire 3d Drivers have been merged, the XFree team reconsidered and continuing with a clear refreshing roadmap. Not to mention that there is NO need to permanently change a project if it's not needed. E.g there is no need to make big changes in the code as long as it's stable and works.

Sure there are a bunch of things that everyone can complain about such as better modularisation with autoconfig system, newer or better separate libraries such as zlib, xcursor, fontconig, freetype2 etc. But after all the majority of people aren't contributing to them anyways except a handful of people. The reasons why Keith and some other people got thrown out is because of diagreement. In a team where you develop with other people you need to follow a bunch of guidelines and agreement with others to achieve a goal. The strict X goal was to be independant to a certain Desktop environment with a good reason because not everyone like to participate to either GNOME, KDE or XFCE and so on. Staying a neutral project was indeed required. But some people who were developing there started to make big noise and thus they got thrown out.

Let's eye over to X Freedesktop for a moment. It's not the ultimate in solutions either. Instead of having one X server they now have ONE X server for each graphic chipset. (As it used to be under XFree 3.x). Do you think this is a solution ? A lot of heavy changes for what ?

We should first see what happens in the next 1 year and not judge about other peoples work or roadmaps. This is the same junk like other people complain about KDE or GNOME while not having contributed anything to it.

Shit happens...
by Anonymous on Wed 25th Feb 2004 10:44 UTC

I guess that's what happens when you ask people to co-operate. Humans aren't made that way - eventaully selfish motivations will lead them to fight and fall out. It's a bit like the collapse of communism I guess - a lot of people say the West won the cold war - but in the end it was more a case of the communist world tearing itself apart. Just like communism it seems that many conflicting voices have led to a hopelessly inneficient development model - with no clear way to turn the project around. Closed development models, personal egos, conflicting objectives, persistent bickering and infighting - ultimately it seems doomed to fail.

As a long term Linux user, I have to say I find the current situation very depressing.


Re: As usual ignorance leads the article
by Luke McCarthy on Wed 25th Feb 2004 10:45 UTC

Just a little note: Y is a complete new design, not a fork of X;-)

by follerec on Wed 25th Feb 2004 10:52 UTC

how about choosing the individual (or two) maintainers from each linux and bsd distribution to become the leaders. someone from freebsd, openbsd, netbsd, redhat, suse, mandrake, and others i can't remember, to become the leaders. maybe they could even decide which parts they like and handle that. i dunno. but it seems better if everyone agrees to help out. of course there might be clashes with "old-and-stable" vs "latest-features" and "bsd" vs "gpl".

[OT]What is the rule for forking regarding the license?
by someone on Wed 25th Feb 2004 11:18 UTC

If I want to start a fork of XFree4.4 have I to continue with their license (this GPL incompatible thing)?

Or can I say for example:"My Xfree4.4 fork is under the LGPL." ?

I think i have to continue with their license because to fork I have to "copy" their source base to my repository.
So I should fork with a version that doesn't have this "evil" license.

Can someone help me?

by Allan Sandfeld on Wed 25th Feb 2004 11:35 UTC

No, you can't change the license freely, but XFree 4.4rc2 was released under the old GPL compatible license, and is thus forkable.

by BlackCat on Wed 25th Feb 2004 12:57 UTC

Looking in my crystal ball, I see the following happening:

Short term, distributions will continue shipping XFree86. Eventually maybe even including the newer versions with the new license, but with F.d.o Xlibs (which is what programs link against) at its core.

The F.d.o X Server will continue gaining momentum and support from the distributions. They'll drag X kicking and screaming into the 21st century, ditch the awful iMake, get rid of the entire /usr/X11R6 tree (what a relic! I read about this on the f.d.o mailing list and it really is going the way of the dodo, Yay ;) . Drivers will also optionally be packaged individually by distributions, so for example you can push xserver-ati-1.2.1.rpm to users with bugfixes without having to upgrade all of X as it is now, and not be tied to the X release schedules. Xouvert's still going nowhere IMO. In the end, the future of X is looking a lot better than it did, say, a year or two ago.

re: As usual ignorance leads the article.
by tc on Wed 25th Feb 2004 13:13 UTC

" there is no need to make big changes in the code as long as it's stable and works."

When a body of code grows so large and so intra-dependent, that minor changes require major rewrites, the time comes for the big changes. Maintainability is a third important consideration. You should always consider both what you have accomplished in the past, and what you may accomplish in the future.

How this relates to XFree86, I haven't got a clue. But you can't leave out that factor.

by Richard Moore on Wed 25th Feb 2004 13:50 UTC

You'll be happy to hear that imake has already bitten the dust in the f.d.o. tree. Rejoice!

Death to X
by Robert on Wed 25th Feb 2004 14:15 UTC

I hope even more trouble will befall X, and that it will finally die alltogether. It is the reason I have never even bothered with Linux (short of the painful things you're forced to do with it at college).

But I guess that by the time a serious X alternative can be released, SkyOS or even OpenBeOS will be long finished.

Definition of X
by Jim-Bob on Wed 25th Feb 2004 14:24 UTC

Regarding the definition of X given by the author, namely "X Server: a piece of software which implements the X11 protocol. This software handles the rendering to the screen for X Clients"

Let it be known that for 20 years now the application running on the host has been called the CLIENT APPLICATION and that any remote display/input device is called an X SERVER. There is NO SUCH THING as an X CLIENT.

Why is it so hard - why is there a refusal - to refer to a device which SERVES up a remote display/input session to a user as an X Server?

v RE: As usual ignorance leads the article.
by FH on Wed 25th Feb 2004 14:31 UTC
New X server hosters/maintainers
by Aanund Austrheim on Wed 25th Feb 2004 14:35 UTC

Here is a thought, OSDL is supposed to have a "task-force" for improving the linux desktop. I think what they should do is make a fork of 4.4-rc2 and invite developers to start developing the "new" "improved" x-package there. A good X implementation is just as important for desktop linux as a good kernel is...

v Testing
by HI on Wed 25th Feb 2004 14:39 UTC
best article of 2004
by the_randymon on Wed 25th Feb 2004 15:00 UTC

Hands down the best article I've read on this site in 2004. Great work, and let's get some more of this kind of stuff, and less of the amateur distro reviews.

Re: Death to X
by sofa king what on Wed 25th Feb 2004 15:24 UTC

Well X does serve its purpose, albeit it sucks for desktop use. Something like would be better, if it only got more support. It has a simple, clean api (xlib is just about the crappiest thing on earth), and for desktop use there is just no point in having multiple clients and one server.
On the other hand something like kde/qt with it's own drivers (like qt embedded) would be just fine (and lightweight too ;)

Grokking the situation...
by Karl on Wed 25th Feb 2004 15:36 UTC

I am going to go ahead and take the plunge and explain what I have grokked concerning the new XFree86 licensing issue.

David Dawes is pissed. He is pissed at Keith Packard. He is is pissed at other ex-XFree86 developers. He is pissed off at Redhat and very pissed off at

Whether or not he has a right to be pissed is another question- a question which none of us can suffciently answer. But this license change is David Dawes' payback- his revenge on the renegade rebellion in the X Windows community which started almost exactly a year ago.

When Keith Packard got kicked tensions which had been around for a long time surfaced. There was a lot of talk about forking XFree86. The folks at and totally downplayed this talk- attempting to soothe now open wounds. released a) Keith Packards new Xserver and b) their very own release of xlibs. This got Dawes very, very pissed. (rhetoric) just who do these guys think they are? now they have the chutzpah to claim that they themselves are now responsible for XFree86.....

In the meantime Dawes acts as if there is nothing wrong with his new license-'if anything needs to be changed its the GPL....'. He is being so unbelievably dishonest. He knows damn well that his changes cause problems with nearly every program which is linked against the xlibs-which is almost everything you see on your X Windows display.

But Dawes failed to grasp what was really going on. He failed to understand that people were not mad at him and XFree86 about the slow pace of development-they were upset because the XFree86 core-group(which no longer exists) was engaging in heavy-handed politics(outsting outstanding contributors) and that a new wlesspring of demand for genuine community participation had arisen which confronted and challenged the old tradition which had dominated XFree86 for years- that they were alone, that no one helped them, that everyone complained but did nothing- that they had to go their own way because no one really supported them.

Now Dawes has his revenge. He simply could not see the demand for a community- the demand from those who had hitherto played only a trivial role in the development process saying that they wanted to see things happen now. Pent-up desire, demand to see changes. Dawes could have endorsed this, he chose to dis it. OF course Dawes is not alone in this-but he is the only one left after the core-group dissolved at whom one can direct their anger. Dawes used to be the more level-headed of the core-group, far mor fair than some of the other members.

Whether you can see the problems with the new license or not is not the issue. The license issue is a thing of "perception"- in more ways than one. The Linux distributors could "sense" that something wasn't right, that something was changing. Now the Linux community and Theo from OpenBSD have made it clear-"NO" to the new license. If they all position themselves saying the new license is "GPL incompatible" the new license *is* incompatible- ie. no one will use or accept the new license. The responsible persons at and have gone out of their way to avoid a fork-yet their actions are, in the eyes of Dawes and others at XFree86, a fork in every sense of the word, except for the name.

The new realities call for new structures and new roles. XFree86 is a staunchly conservative organization-which for the most part has been a Good Thing(TM)-it has given us a reliable, stable and free X Windows implementation for many years. Change is in the air and XFree86 has found itself incapable of dealing with the new situation.

Luckily the folks at are really committed to the future of X-they know that X is not simply a Linux thing, and not simply an x86 thing-they respect and endorse the standards, the protocol, and in so doing are remaining loyal to that which XFree86 has stood for. has only stepped forward due to the fact that XFree86 has been unwilling to really open up it's development process. If XFree86 would have simply reliquished it's heavy-handed control of development this whole issue would have never arisen.

The license issue is only making apparent what is and has been going on behind the scenes. Re-Changing the license back to its former state, or changing the GPL will not solve these issues. If XFree86 chose to re-change(ie.revert back) to their original license it would signal a concilliatory move-it would be a gesture of good will. But wounds have already taken place. There is need for much healing. Trust has been violated.

I am not saying that Dawes is the only or main culprit here. Many ego's have been involved in this mess. Remeber it always takes two to tango- their are legitimate points of view from all sides involved. As it stands XFree86 has backed itself into a corner and is now becoming the outsider. If they continue this path they will become irrelevant.

I wish all the parties involved could just talk things out and agree to new structures and new roles. I doubt this is going to happen- the attitude now in circulation is- "fine- fuck you, if you are going to act this way -well we don't need you". This is horribly divisive-it is tearing the threads which hold *the* community together apart. Right now the main battle is about the identity of this- *the* community.
Such situations can galvanize new identities-it can also lead to splintered groups working against each other. Only time will tell how things will develop....

Re: Death to X
by Rayiner Hashem on Wed 25th Feb 2004 15:39 UTC

C'mon, this is getting old. X is pretty damn kick-ass. Programmed properly, its quite fast too. As for lightweight --- you do realize that it actually takes more resources to make something seem faster? For example, if you want to eliminate redraw during resize, or redraw after expose, you need to double-buffer windows, which takes up an enormous amount of memory.

The FD.O X server will have tons of cool features like OpenGL acceleration, double-buffered windows, etc. And all of it will be done within the X framework!

RE:As usual ignorance leads the article.
by Chris on Wed 25th Feb 2004 16:22 UTC

Once again. Great article, this is actually news without a bunch of interjected editorial tripe. If everyone in the media could produce articles like this the world would be a better place.

by Howie S on Wed 25th Feb 2004 16:46 UTC

It appears from the screenshots that Keith Packard's X server is very 'buggy'

(Sorry, I just couldn't resist.)

Good summary indeed.
by dpi on Wed 25th Feb 2004 17:10 UTC

On FOSDEM there have been a lot of talks about general GUI (ie. FreSco), hardware + GUI (ie. Robert Love's speech), and the future (DirectFB, FD.o). Keith Packard held a speech which basically was about the same as yours, except you have a lot of references instead which is quite hard to do in a RL speech. There's a small difference regarding the future of FD.o (and X? ;) though.

First, there are basically as what i see as 3 streams who both will go their own way. Whoever becomes mainstream in which timespan is what we don't know yet...

1) A conservative group who do not use radical solutions.
2) A progressive group who build upon X.
3) A progressive group who code from scratch.

1) The people who think X (XFree86) is now good enought and who keep thriving on this. For example Enlightenment has a hack called Imblib(2) which works around XFree86's design, snapshotting a background, using that in a different window. This is an ugly and slow hack, but it works more or less. There will be people who will, for the time being, develop their software around these hacks and other hacks. How many and on which X server / implementation is also in some extend in the hands of mr. Dawes & co.

2) There are people who base their code upon X (XFree86) and build upon that ("radical") new things and people who base their code upon this. The GNOME people for example cooperate with FD.o. FD.o is also the only example i can honestly name and i'm not sure wether Xouvert and FreSco fall under group 1 or 2.

3) There are these people who from scratch code a solution to the current problems. Y (which has as goal backwards compatibility) is an example, DirectFB (which is backward compatible with X) is yet another, and there are probably even more (other FB implementations?).

Whatever is going to succeed it has to be better than the current implementations and when they want a clean change, backward compatibility OR cooperation with tons of GUI programmers to call for a change, is unevadable. The latter is extremely difficult and lots of work, so going for the former seems the best choice.

What Keith kinda pointed out as a difference is how he sees FD.o's future -- specifically the code which has not yet been coded. If i understood correctly he wants FD.o to be extremely modular 3D environment with the X server only being a GL server.

XFree86 License
by Inglorion on Wed 25th Feb 2004 17:34 UTC

Note that, contrary to what the article states, the new XFree86 license does NOT have an advertising clause. It just requires one to acknowledge the work of the XFree86 team, in the documentation or in the program, wherever other such acknowledgments appear.

This is not anywhere near as outrageous as some people make it out to be, and may not even lead to GPL incompatibility.

by Shawn on Wed 25th Feb 2004 17:41 UTC


Note that, contrary to what the article states, the new XFree86 license does NOT have an advertising clause. It just requires one to acknowledge the work of the XFree86 team, in the documentation or in the program, wherever other such acknowledgments appear.

It may not be *exactly* an advertising clause, but it's darn near close to one.

by Rayiner Hashem on Wed 25th Feb 2004 17:44 UTC

If i understood correctly he wants FD.o to be extremely modular 3D environment with the X server only being a GL server.
Actually, it'll be slightly different. The current approach seems to be leaning towards having the GL drivers be in an independent driver layer, and the FD.o server sitting on top.

I also don't think they're moving to a 3D environment, but rather, a very rich 2D one where advanced graphics are made possible by a vector-graphics API (Cario) accelerated by the underlying GL mechanism.

Well done
by z1xq on Wed 25th Feb 2004 17:52 UTC

A very good article.It educated me about many aspects of the XFree86 project. The way I see it a major shakeup in an open source project is occasionally needed to cause a paradigm shift. Out of conflict comes change and innovation, though it is seldom a painless process.

Re: XFree86 License
by johnlein on Wed 25th Feb 2004 17:59 UTC

Its not just the GPL compatibility.
Theo DeRant wasnt to hapy with the new license and he is no GPL proponent, Dawes response to everything at the moment is "you are full of shi#".

XKDE / XGnome
by Savage Sailor on Wed 25th Feb 2004 19:11 UTC

I wonder why KDE or Gnome haven't expanded their scope and built down to make an XGnome for instance. It would probably make an OpenGL accelerated desktop easier to deal with.

@Rayiner Hashem
by Sagres on Wed 25th Feb 2004 19:17 UTC

"Actually, it'll be slightly different. The current approach seems to be leaning towards having the GL drivers be in an independent driver layer, and the FD.o server sitting on top."

Does that mean something else can sit on top of that layer, other than If yes it's good news

OpenGL and Linux.
by cheezwog on Wed 25th Feb 2004 19:41 UTC

What worries me about 3d graphics as part of the desktop is that there are virtually no open source drivers for hardware acclerated 3D.

I think some 3dfx boards and a Matrox card have free drivers, but that's about it.
Therefore, it's a little dangerous moving towards a desktop that will only ever work smoothly as part of a non free system.

I think it's important to have free open source graphics card drivers as I have found propriatory ones to cause endless problems on Linux . ie, I can't build the kernel I want as breaks them, they are unreliable compared to the xfree ones, and they give a time limit to the useful life of your hardware (They will only keep up to date with the Linux kernel for a limited time).

I hope a solution is found, as graphics cards are becoming so complex that even with specs, writing an open source driver is a huge task.

by Rayiner Hashem on Wed 25th Feb 2004 19:54 UTC

@cheezwog: Let me correct myself. The driver layer, as currently envisioned, won't just be for OpenGL. It'll be more akin to what XAA + DRI is now, only abstracted out of the X server itself. So the X server will be able to fall back to software rendering or regular 2D rendering if an OpenGL driver is not available.

@Sagres: In theory, I guess. Certainly, it'll be easier to use it than to load current XAA drivers on a non-X window system!

Community is preparing to purge
by blahblah on Wed 25th Feb 2004 20:15 UTC

you can smell it, this license issue is not the central issue, but its a whiff of the real conflict.

this is healthy - its become clear that as group is not functioning. the community is purging them before they infect the rest of the ecosystem. i for one am heartened by the fact that some notion of market economics (even if the currency is goodwill) is functioning in the open source world.

by marc on Wed 25th Feb 2004 20:28 UTC

The issues with XFree86 are bad. Changing the license was the tip of the iceberg and the community has united to solve the X server problem. XFree86 will become irrelevant so Mr. Dawes can look for a slug as of now...
He'll not change the license as it was and his project will become isolated. The OSS community can't be forced into this by 1 man. If he wants to change the GPL then he's acting wrong. Mr. Dawes, at least in the USA you say "PLEASE" before you try anything else... Forcing the community won't help, but merely make allot of people hate you. You're a smart man so act smart, you are intelligent so act intelligent, you work hard so I think that by your move you're just wasting your work...and time...if you don't care anymore about X then go somewhere where you can make more money and leave the community alone... Or has the community left you? lol
Just my 0.00002 pennies...

I hope developers will be more focused
by hyko_l on Wed 25th Feb 2004 22:06 UTC

It's not very hard to anticipate that in a couple of years, 3D desktops will become quite usual for Windows users.
It would be a shame if by that time, Linux users would have only plain old X server as an option...
I can only hope, developers behind xfree86 and other free X server implementantions will join efforts and become more focused, building a healthy Xserver which can evolve and match the expectations of the future desktops.

Re: XOuvert
by Deb Ian on Wed 25th Feb 2004 22:32 UTC

<quote> XOuvert is not a fork but consider themselves as branch (just visit their webpage). And yes, together with will fork XFree86 starting with 4.4 RC2.</quote>

Where did you get the information about a fork?

3D Desktop
by Savage Sailor on Wed 25th Feb 2004 22:34 UTC

It really is a mess on the 3D side. I'm developing a cross platform Open Source game. It takes twice as many debuggers and testers to make the thing work with Linux than with Windows. We need to grow a chicken in a petri dish so we can avoid the chicken/egg conundrum. The chicken is either open source hardware or a 3D "killer app" that only works with Linux.

Great Summary
by Topher Cyll on Wed 25th Feb 2004 22:52 UTC

This is a great resource for bringing people up to date on the free X11 situation. Very well done!

by Anonymous on Thu 26th Feb 2004 01:08 UTC

All I can say its been nearly 10 years and Xfree cant still clear the video memory before it goes into graphics mode....

by Rayiner Hashem on Thu 26th Feb 2004 01:25 UTC

What on earth are you talking about? I get a nice little NVIDIA splash screen when I boot up X ;)

the new "Linus"
by foo on Thu 26th Feb 2004 03:15 UTC

Regardless of whomever may step up to the plate as the author of the article calls out for...lets hope that by the time the next big closed source desktop is ready (1? 2? 3 yrs from now?) that the Open Source's foundation for GUI's will be *so good* it's scary!

The possibility that the "new leader" may never emerge always exists. Rather a collection of assorted efforts from all different paths may come together and turn the tide...and that is never a wild prediction in this trade ;)

I haven't been in Open Source all that long, but I am still quite surprised at all the hostility toward the XFree project...well deserved as it may seem these days...

foo out

Excellent Article
by nk on Thu 26th Feb 2004 09:40 UTC

one of the best OSnews article so far. well done

re: Xfree
by sofa king what on Thu 26th Feb 2004 09:51 UTC

And it's 2004 and it still doesnt detect and configure plug & play monitors like it should.

Re: Xfree
by hive on Thu 26th Feb 2004 10:53 UTC

And it's 2004 and it still does not have proportional mouse acceleration, only this threshold jump.

@ Rayiner
by dpi on Thu 26th Feb 2004 15:29 UTC

Thanks for the clarification. Do you happen to have a good explanation/summary about this? I couldn't find the slides of FOSDEM on the mainpage.

by Rayiner Hashem on Thu 26th Feb 2004 21:12 UTC

I don't remember anything on FOSDEM about this. It was mostly in the mailing lists. Try reading the xserver lists for 2003-November and 2003-December to the threads "new device driver features" "dri integration" and "xserver on OpenGL". There is a sub-thread on 2003-December called "mesa-solo status" that is pretty informative. Hope that helps ;)

Configurable X
by Dan G on Fri 27th Feb 2004 01:10 UTC

I would hope that while X achieves a more open development structure, it would also be as configurable as the linux kernel is - prior to being compiled. That way, everyone can have their code and features included, but not everyone would have to run the code they dont want/need. Is this what f.d.o means by more modularized?

Of cource this would require the current X project leaders to get over their power trips/egos.. not likely

RE: Xfree
by CF on Fri 27th Feb 2004 14:53 UTC

And it still does not have an easy way for users to install and use smart vector fonts (OpenType, AAT, Graphite) necessary for languages using complex scripts.