From Slashdot: “Keith Packard wants to fork the XFree86 effort. […] The XFree86 team is trying to become more open, to combat the fork. Keith is a capable developer, having worked on FontConfig, Xft, the X render extension etc. Meanwhile, All is not good in how XFree86 drivers are being developed.”
Possible XFree86 Fork; The Organization Asks for Input
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
What are the ideas behind the proposed fork?
I am shocked by this incident and the way it is being handled by the Xfree86 team.
And whatever Keith is trying to do, first he should get a chance to defend himself.
Apart from that, the more reasonable comments I saw on /. were by Alan Cox, who thinks that forking XFree86 may not be such a bad idea after all.
I haven’t seen detailed info on the urls posted, on what reasons keith has behind such fork, anyone know where to get more info?
without that info i would venture to say it’s a bad thing to fork such a massive project since it’ll make developers jump from one to the other, thinnin the resources available, atleast that’s how i see it.
After reading the “Diary for mharris” (link on /.) I think that a fork is overdue. XFree86 seems to be very poorly managed (if the mharris diary is accurate). Especially the comment that ATI patches need to cook 9 month till they are applied and the comment about “we are no longer accepting applications at this time” were quite shocking for me.
BTW, is Keith Packard still working for SuSE, or do I confuse this issue?
Check this:
http://www.xfree86.org/pipermail/forum/2003-March/000001.html
The tone reminds me of the purges within the Communist party during the Stalin regime, with just a few words changed:
It has been brought to the attention of the Central Committe that one of its members, Keith Packard, has been actively (but privately) seeking out support for a separate Party that would be led by himself. He is also in the process of forming a by-invitation-only group of vested
interests to discuss privately concerns he has about our Country and the future of Communism. He has consistently refused to even disclose these concerns within the context of the Central Committe, which makes his membership of that Committe unviable. As a consequence, Keith Packard is no longer a
member of the Central Committe.
Body found in a dump the following week with a bullet in the neck…
Well first of all I agree that Xfree really needs some fresh air. Specially I would like to see autoconfigure and flags to choose the configuration the way I like to see without tweaking host.def on the otherhand from personal experience from checking out CVS the past 1-2 years now in a regular base I must say that CVS is regulary worked on. A lot of things got fixed, changed and so on.
On the otherhand I ask myself if it doesn’t make more sense switching to directfb.
oGalaxy, No, it doesn’t make sense to go to DirectFB, for the same reason why other OSes have problems: no drivers. And when there are no drivers for the number of gfx cards XFree supports, there is no market.
Anyways, I believe that the XFree organization has a problem. The guys who created it want it to be tightly controlled and that is GOOD, *but* they are too much into legacy and they are afraid of new things. Remember the interview I did with the founder of XFree a few months ago? I asked him about adding some cool new stuff on xfree that can be found on BeOS Dano and MacOSX and he seemed quite conservative. And if a fork will help bring new stuff to XFree, then a fork it is.
If you manage to contact Keith and get some straightforward answers above the present /. noise, I think that would be great.
Also, can you post a link to your interview that you mention above?
Thanks,
Andrew
I must say that after reading it, i must agree with eugenia and andreas. for those who want to read it :
http://www.advogato.org/person/mharris/
And breaking legacy every once in a while might be a great thing
, where would windows be if MS hadn’t decided to drop DOS… bad example i know, but works, where would intel proc’s be had they not kept all that uneeded stuff from 8088,8086 time… even closer to being risc than they are now.
http://www.osnews.com/story.php?news_id=1294
People complain that X (and XFree86 in particular) sucks. There is no clear roadmap for future X development. Maybe this will give the whiners a chance to help out.
A big question is whether the forked XFree86 will use the same license! I think XFree86 uses the MIT license, which is equivalent to the BSD license without the advertising clause. So the new fork could be GPL or even closed source!
You’re quite right about DirectFB lacking drivers, maybe if they made an efford to use scitech SNAP things would change a bit. Another problem is that currently DirectFB only works with GTK and doesn’t support QT, they would need more developers for that i guess. If this situation changed we would probably start having DirectFB based distributions.
That however doesn’t change the fact that DirectFB is a more modern and desktop oriented system than X IMHO.
Well look at that embedded QT for PocketPC’s they don’t even need a bottom X layer at all.
I know it is a huge project and a ton has been invested in it, but frankly it is old and bloated. Given the clear organizational problems in addition, I think a fork is long overdue.
Maybe Eugenia can throw open the discussion, or get some interviews going on what the fork would look like – perhaps this is a chance to not only fix the problems with XFree86 development organization but fix XFree86 itself.
How ’bout it Eugenia?
there is a ton of cruft that can be removed, a ton of layers that can be removed(right now, there are far far far to many layers to pass through to get from the server to the client especialy in todays GFX cards), and a lot of additions that will make performence better can be added. heck, they could even make it a possability to disable the network transparencey if the user wishes.
runtime wrote:
> So the new fork could be GPL or even closed source!
Hmm… note that XFree86 is not a GNU project. What could really garner support for the fork would be to GPL it and get the FSF to make it an official part of GNU.
> You’re quite right about DirectFB lacking drivers, maybe
> if they made an efford to use scitech SNAP things would
> change a bit […]
please. How can anyone seriously consider the framebuffer, with the kind of video hardware we have now? it’s not uncommon to see people using video cards with one time and a half the CPU of their machine, and many times the RAM (how long before we see cards with their own hard disk?). I mean, people, there’s cards dedicating a whole FPU to a single opcode
Even something as trivial as bit block transfer is better done by the card itself, rather than with mmap+memcpy: it takes zero CPU, and allows the card to have larger words, and handle more memory, than the machine that hosts it
DirectFB has no drivers because it’s meant for machines without a real video hardware. X11 scales better, in that regard. Too bad it has an out-of-process server design, which, coupled with the stream-oriented protocol, dooms it to always be somewhat unresponsive
There’s no need to fork it then GPL it.
That would be the dumbest thing in the world.
Right, so then the GNU Project could benefit from the years and years of development that when into XFree86 just so they could turn around and relicense it in a way that the original project could not benefit from the GNU changes? Oh, but wouldn’t that be a pleasant irony?
GPL Community: “Look at all the evil corporations who are stealing your MIT/BSDL code and not sharing their changes with you! You must use the GPL or else Bad Billy G. Will steal all your code! Evil. Evil.”
Later on when nobody is looking….
*GPL Community steals code and renders it useless to original project*
Can anyone say hypocracy?
Don’t give me that hogwash how the MIT/BSDL encourage everyone to steal the code. That’s irrelevant to the GPL Community’s hypocracy.
If anyone could really pull off a successful fork of XFree it would be Keith Packard. Normally, I am against forks (I am also against duplication of projects) but this might turn out to be an exciting project.
Still, I follow the mailing lists and it looks like things have gotten better as of late. The developers have been getting a lot better about answering questions for the users, and the community is starting to build up.
performance issues alone should have clued in the Xfree devels about 15 years ago. Why has this taken so long? Will it help? Can anybody, no matter how skilled, actually remove the cruft without removing Xfree entirely? Just looking at the design, it appears that Xfree is a boat whose entire body is built of patches; remove them, and you’ve got nothing left. The driver model is archaic, latency issues are a perennial problem, and interfacing with X is annoying because of poor API design.
Why not just work on OpenBeOS instead?
Dear Mark Swaffer,
No one said anything about GNU “stealing” code from Xfree86. The original code would still be available to everyone, the only thing that would be GPL would be GNU’s changes to the code. Besides, you missed the point of the original poster. Their point seemed to be more one of building an alternative X11R6 server into the GNU project would really provide a boost to both the GNU project and to X in general. For pretty much any software project and/or protocol competition is a good thing.
THE most important argument for still installing windows to me is that with windows, hardware manufacturers can build their own drivers, that will work as a binary with my OS, without my having to recompile the graphics layer of my system.
a good driver model that does not rely on kernel point releases would be super.
Bruno.
Ok, a lotta you seems to think that X is slow, why? Because it has a networking layer…
And that s completly wrong : when using in local, X is using kernel socks instead of tcpip ones so the netstack isn t the reason of the non-responsivity of X :
try running X with DISPLAY envvar set to 127.0.0.1:.0 then try with DISPLAY=.0 (the default), you ll see a big difference in term of performance! In the first case, X uses tcp/ip, with .0 it uses kernel socks!
CU
what a stupid idea to change the license to GPL. some of the GNU hardliners are really sick!
Fresco (http://www.fresco.org/)? Anybody?
DirectFB has no drivers because it’s meant for machines without a real video hardware.
Actually no: http://www.directfb.org/modules.xml
They don’t have enough drivers because they have to code them themselves and lack the documentation to do it.
Utilizing Scitech SNAP in Linux – This idea could bring some real stability to Linux in terms of modern graphics support while also ensuring backward compatibility (SciTech’s API is rock solid in this regard) a concept that xfree86 version 5 is looking to abandon. Additionally HW vendors might also look favorably upon this concept as SciTech has long standing relationships with a majority of the manufacturers.
This may help in understanding the situation.
http://www.xfree86.org/pipermail/forum/2003-March/000128.html
The final comment on client-server displays irrelevance is interesting.
I’m the lead developer of a ‘competing’ system to X, and yet I always end up having to defend it…
<p>
Client/server is not evil, it is a great way to separate components of a large system in flexible and usable way. Just because you don’t run X over the network doesn’t mean the authors of X should use a different design methodology.
<p>
DirectFB is not a replacement for X in the general case. It’s good for set-top-boxes or other environments where you generally only need one application, and you’re mainly playing video. DirectFB does not have the range of acceleration support that X does, and it’s designed to run a single application.
<p>
The problem with X isn’t the design, it’s the implementation. I’m not talking about speed, I’m talking about modularity. You can’t take XFree86’s protocol layer and stick it on top of a completely different rendering system, or take X’s drivers out and use them in http://www.ggi-project.org/“>GGI Fresco” rel=”nofollow”>http://fresco.org”>Fresco, or <a href=”http://picogui.org“>PicoGUI.
<p>
So why aren’t the X developers rearchitecting X so that instead of being a monolithic architecture full of misconceptions, it can be a system for powering whatever GUI the user needs, and experimenting with new designs? I’ve asked this, and the only response I received was that the entire world should run X.
<p>
I really hope Keith goes ahead with this. X11 is definitely one of the largest stumbling blocks in Linux adoption as a mainstream desktop OS, and it now becomes apparent why this is the case: close-mindedness and the desire to control “intellectual property” instead of the quick adoption of new ideas. Keith has been responsible for many of the most needed changes to XFree86 lately, including fontconfig. Why do you think he is taking this radical step? It’s obvious that working with the core team is no longer possible.
The promise to become more open rings hollow with the quick decision to exclude Keith. This could well be XFree’s suicide note. Hopefully those who want a new, more open, more capable X11 to prosper will join Keith.
Is that Xfree86 has ALOT of extensions, however, they’re not used by desktop developers because of many of these extensions are not available on XFree86 ports on other platforms such as NetBSD and DRI, or, these extensions have not been merged into the X tree meaning that if one were to create a desktop for all X servers, one would have to stick with the X specs which are always behind XFree86 in terms of features.
What there needs to be is faster merging of XFree86 changes into the main X specification so that all the commercial X servers are kept in sync with Xfree86.
Good news. XFree is being developed with a lot of features that 90 % of users don’t know or will never use.
Can’t disagree about the attitude of the XFree Core Team if they saying exactly what happened.
But a fork to make XFree with less code to load on ‘startx’…
and an attempt to make it more lightweight (who really uses the network feature ? Can’t they remove it and create a ‘Desktop source code version ?) and with a new config file that’s more straigthforward to a person with average computer knowledge and average brains can only open a good road for the future, in my modest opinion.
Forking is not always a good thing neither is it bad. In this case though I think it is NEEDED. Not undermining the significance of the Xfree86 project, IMO it needs some restructuring and a much clearer roadmap. The way how things work currently is just too slow for the needs of the fast paced development of projects in unix based systems (i.e. desktop environments, etc).
Someone posted about the FRESCO project… indeed the concept of the project is promising, in fact they need developers and support. It is unclear though how this project pans out in the current problems Xfree86 is experiencing.
–just my opinion–
who really uses the network feature ?
I do! A cheap laptop makes a great thin client to my desktop machine. After sitting at a desk for 7.5 hours a day at work, it’s nice to be able to use my desktop remotely from the couch.
I know the X standard supports both threaded and non threaded X-server’s, what’s the situation regarding XFree86 when i run top or “ps -e” it always appears as one process. so it appers to be non-threaded. If this fork goes ahead i think that producing a multi-threaded X-server should be one of the goals
Keith has posted an email called “A Call For Open Governance Of X Development” it can be found here:
http://www.xfree86.org/pipermail/forum/2003-March/000168.html
Well, it is good to hear from Keith Packard. It is very difficult to see him voice out anything before. Hopefully he will continue contribute his good work to the X community.
Forks can be good, there are zillions of examples.
Forks can be bad, there are zillions of examples.
The real problem is that, quoting Alan Cox, “The response [of the XF86 Core Team] to Keith is horribly netbsd’ish though, this “you are with us or against us” thing (Actually its terribly George Bush right now).”. The only loser is the XFree86 Core Team. You say that you want your development to be open, and start… bombing maybe your most talented devel!
And about DirectFB… I don’t believe in DirectFB playing a role unless it runs on the same range of Unix software that X runs. A Linux-only solution is bad.
Andrea Cisternino: The final comment [of the AOL guy] on client-server displays irrelevance is interesting.
chicobaud: who really uses the network feature ?
Network transparency is an invaluable tool in corporate environments, and especially in the Third World.
Here in my work, we still can use P100/32M machines to run giant things like Mozilla, thanks to the $DISPLAY magic. It lowers dramatically your TCO because you needn’t to maintain a lot of installations, only a small number of servers.
For desktop users, maybe it’s irrelevant. Unless you start to having two, three etc computers at home.
who really uses the network feature ?
Client/server isn’t a feature, it’s a design technique- you could of course rip out all the network-related code from X and link the rest into your toolkit and app, but then you’d have a mess that would only run one app and negligibly faster than X running locally.
Client/server doesn’t need to be done away with, just rethought a little. Both Fresco ( http://fresco.org ) and PicoGUI ( http://picogui.org ) are client/server, but they use new architectures intending to circumvent the problems with speed and inconsistency that X has.
I think most of those core developers are missing the Tao..
To me it looks like they’re backseat developers now, who kicked one of the best developers out of the ‘core team’, who made the most promising contributions to XFree.
I would vote for kicking out most of those core developers instead.
(the ones who can only talk about political issues but don’t contribute anything anymore, at least, I haven’t seen them contributing much)
Interestingly, almost all of those extensions are ignored because they turned out to not be such a good idea… The extensions that are high quality (GLX, RENDER, SHM) are supported by most implementations.