XFree86 4.3.0 Released

Dunkalis writes "The latest version of XFree86, 4.3.0, has been released! Release notes here, mirrors here. Enhancements include drivers for newer Radeons, better PS/2 protocol detection, the XRandR extension, better font support, and more!" Source tarballs are available, or wait for your distribution to package them...

Comments

Re:Still buggy...

[djcapelis]

Heh, it's a point-oh. Give it time, they'll work out all the rougher spots. I'm gonna wait for the distro to pick it up, they'll make sure it's stable enough. Though I'm dying for the new i810 support... tuxracer is unbearably slow without it.

Re:distro release

[silvaran]

Red Hat 8.1 beta (Phoebe) has xfree86 4.2.99.3 packaged... since then, XFree86.org has released several more snapshots (.4, .901 and .902)... I've been running the snapshots (.3 and most recently .902) for awhile now... .3 had a problem with the nvidia driver... once X came up, I couldn't Ctrl+Alt+F# to a terminal, but that was fixed fairly quickly.

Anyways, RH is likely waiting to test all these newfangled toys. GNOME 2.2 came out, and now that X4.3 is out, RH8.1 shouldn't be too far behind :). Be patient! ;)

Let the flames begin ... and ignore them.

[IGnatius T Foobar]

I am a happy X user.

Since this is a story about X, all of the pre-programmed Slashbots are going to trot out and declare that X is broken, old, badly designed, missing features, whatever.

Meanwhile, the XFree86 team continues, release after release, to pound out great code that addresses all of the shortcomings people tend to cite. Faster direct rendering? Check. Anti-aliased text? Check. Multi-head? Check. Video extensions? Check. 3-D? Check.

Do you see a pattern here? X is versatile. X is extensible. X is the industry standard -- all Unix GUI programs use it.

And as always, X's killer feature is its network transparency. No "desktop-within-a-desktop" nonsense like you have to do on other platforms. Today I had the windows of programs from no less than three different computers running on my desktop. Transparently. Lots of X users do this every day, usually without even thinking about it.

Perhaps someday the tired old "X is obsolete and must be replaced" will finally cease. But today is probably not that day. Let the flames begin. I will ignore them and continue to praise the XFree86 developers for another job well done.

Re:Let the flames begin ... and ignore them.

[starseeker]

Usually the replacement mentioned is Fresco, a.k.a Berlin. I think Fresco is a good replacement for X in the same way GNU Hurd is a good replacement for Linux. The dieas and potential are true next generation, but the implimentation is years away. Which is fine. I agree X certainly will hold us as long as it needs to.

Re:Let the flames begin ... and ignore them.

[captaineo]

All Unix GUI programs use X for the same reasons that (almost) all businesses use Microsoft Word...

Re:Let the flames begin ... and ignore them.

[kcbrown]

Anti-aliased text? Check.

Uh, no.

There's not really any more support for anti-aliased fonts now than there was before the Render extension, except that the Render extension makes drawing of anti-aliased fonts fast.

But the application still has to do the work itself. Whether it does so through "standard" libraries like the gtk or KDE toolkits is irrelevant: the bottom line is that anti-aliased fonts requires client-side support.

Clients have always been able to do anti-aliased fonts if they wanted to, but prior to the Render extension they had to do it the hard way: by actually drawing the individual characters and doing the transparency blending themselves.

The implementation of anti-aliased fonts is all wrong, IMO. The XFree86 folks should define a new font server protocol that knows how to talk about transparency (indeed, the protocol could easily be implemented such that it uses the same socket and everything: if the X server sees that it's talking to an old font server then it will revert to the proper monochrome font protocol), implement a font server based on Freetype that actually uses it, and hack the X server backend so that it automatically does the right thing when asked to perform operations using an antialiased font. The client shouldn't even know or care if the font is antialiased: that's a server-side-only issue (it's acceptable to name antialiased fonts differently, using perhaps a different encoding name or something, in order to make it possible for the client to distinguish between antialiased and nonantialiased if necessary).

Font handling belongs in the server. The client should never have to worry about it. Which means that the situation as it is now should never have come to pass. The Render extension is very useful for things like doing transparent windows and such, but it should never have been used for font handling: that was an evil hack, and now we're stuck with it.

Re:Let the flames begin ... and ignore them.

[Tyreth]

What's this all about? I'm not biased towards or against X, I simply don't know enough.

But I do know that Microsoft also improves Windows with each release, addressing many major complaints. We still don't support them despite all this! So I don't see how your argument is useful at all..

And besides this, yours is the first modded up comment about X and whether it's obsolete or not.

Volunteer, or Donate

[Anonymous Coward]

Debian is all volunteer efforts. Why not help them out, after all they've helped you out plenty. Then volunteer your time/efforts and compile yourself and make a package for others to use.

Or donate $$$ to the Debian project.

What goes around comes around.

Re:Still buggy...

[lspd]

I'm gonna wait for the distro to pick it up, they'll make sure it's stable enough.

Depends which distros your talking about.. Some seem to care more about buzzwords than stability. But there are a few exceptions.

Personally, I wouldn't touch it with a 10 foot pole.. I don't pretend that I can help debug it in a meaningfull way...and this new version doesn't make my current version work one bit less.

Stability issues aside though, I'm overjoyed to hear that Radeon support is still improving. I'll pass a brick when DRI + Xinerama works with the OS Radeon drivers. Improving support for built in 3D chipsets is also great news. Even minimal performance is a godsend. These guys are doing great work.

Windows/OS X architecture is similar to X11

[g4dget]

My overall performance isnt worth "network transparency".

Well, so tell us: in what way are Windows or Macintosh OS X supposed to be more efficient? Where are these great gains in efficiency in their architecture supposed to come from? I mean, it can't be the use of IPC or system calls for the application to communicate with a graphics server: Windows and Macintosh have that as well.

In reality, there is no fundamental difference in the client/server window system architecture between OS X and Linux. For NT, there is a difference: large chunks of the windowing code have moved into the kernel ad some point, but you still need system calls to talk to it. Of course, there is nothing to stop anybody from moving X11 into the kernel.

Overall, the idea that network transparency is some sort of special feature that one pays a high price for is nonsense: all major desktop operating systems run in protected mode, and most GUI applications run in a different context from the window system. X11 simply has been designed that way from the ground up, while Windows and Macintosh have evolved there from "direct mode" graphics. Network transparency in X11 is not so much an issue of IPC or how it does graphics--it uses IPC like all desktop windowing systems--but in having well-defined network transparent support for features like window management and configuration information. It's lack of those features in Windows and OS X that means that Windows and OS X are not network transparent.

In practice, XFree86 is a damned efficient window system that, when it has comparable drivers for the graphics cards, beats OS X handily in terms of performance and memory usage, and usually even beats Windows.

You need screen on another computer, use TightVNC.

TightVNC gives you a "screen on another computer". It does not give you network transparent windowing. If you are running a well-designed X11 desktop, you can run applications on any machine, and they will behave as if run locally. You can also move individual windows between machines and displays. Of course, Gnome and KDE both break this behavior, but that's not X11's fault.

MSWindows 98 is snappy, even on quite old hardware. XFree runs like shit. It feels klunky and laggy.

That's a ludicrous claim. X11 worked reasonably well on 1988 hardware already. X11 servers obviously can run like a charm on 1998 hardware, hardware that's more than an order of magnitude faster.

And that's also what one finds in practice: Windows 98 requires much more hardware (memory, CPU power) to run than Linux/XFree86. If you claim were having a problem with Linux/XFree86, either you are making it up, or you had a bad driver, or you misconfigured something.

Re:Windows/OS X architecture is similar to X11

[g4dget]

I haven't much experience with OSX, but with Windows... KDE/Gnome/Windowmaker all feel klunky with dealing with any type of object (window, icon, pointer). [...] And even with the (older and) newer Xfree86'es, I STILL have the "page tear effect" in many applications. Even in Mozilla. I'm sure I'm not the only one (and if I am, how do I fix that)?

So, KDE/Gnome/Mozilla are clunky. Let's add Java to the mix, too. You won't get any argument from me there. Those toolkits rely heavily and unnecessarily on bitmaps. They are effectively written with a local direct drawing model in mind. They fail to use the X11 APIs properly in many ways. And that's not really so surprising: KDE, Mozilla, and Java all use cross-platform toolkits, and they have been designed primarily around Microsoft's APIs and Microsoft's performance characteristics, with X11 support kind of as an afterthought.

I also know that X wastes a bunch on bandwidth that Tight VNX saves. Try that "oh-so-nice-network-transparency" over a modem. I have, and it sucks compared to how snappy a 640x480 8bit black background screen transfers over EVEN regular VNC.

The default X11 protocol is optimized for minimizing CPU usage and latencies assuming LAN-speed connections. That's the environment people are using it in, and that's the environment it has been successful in for 20 years.

X11 does not try to minimize network traffic. If you want to run X11 over slow connections, you need an X protocol compressor. One comes built into your X11 server, but you need to enable it (lbx).

And there is nothing wrong with using VNC--it's a great system. It simply isn't a network transparent window system, it's a remote display--different function and different application.

still stand by what I said. Go get a copy of WIn98, and a feature-equilavalent copy of Linux with X and managers.

Well, and I say: you are wrong. I have run X11 and Windows 95 on the same hardware, I have run X11 and Windows 98 on the same hardware, and I'm running X11 and Windows XP on the same hardware. X11 has always been competitive with Windows, and usually beat it, in actual measurements as well as "feel".

Still, why didnt you approach the 3-D issue? 3-d's dog slow, even on supported hardware (eg: nVidia).

What's there to address? 3D games under X11 don't involve the X11 protocol, they use DRI (the equivalent of Direct3D). If that is slow, it's a problem with the 3D drivers or the game, it has nothing to do with X11. Basically, this claim is characteristic of your reasoning: something doesn't work, and you blame X11. I have to say, I have run all versions of Quake on Linux and have had comparable frame rates to running it under Windows.

But to what you accuse me of, it's MY fault X runs slow....

It's your fault to make bogus comparisons. Face it, Linux is still largely unsupported, and it is certainly unsupported on Windows 98-style hardware. It's not suprising that you might configure your machine less than optimally or that your drivers aren't very good. Most people aren't bothered by that, since it runs fast enough, but if you are going to nit-pick about performance, you have to nit-pick about your own installation as well.

Likewise, a lot of GUI developers (KDE, Gnome, etc.) come to X11 without knowing much about X11 or understanding how it works. It isn't surprising that they produce inefficient or sluggish code.

But we know X11 can run fast. You said so yourself: you found it responsive on a 120MHz SPARC, hardware that it is actually supported on.

If you want X11 to run fast under Linux, either figure out how to configure it properly and buy the right hardware for it, or go out and pay the money for a commercially supported version with drivers that were written with access to closed hardware specs.

Re:Still buggy...

[The_Dougster]

Yeah, my general policy with X is "if it ain't broke then don't fix it!"

This applies to major versions, like upgrading from 3.3.6 might be a good idea, but 4.1.x might not, especially if 4.1.x works good for you.

Sometimes you have to stick with an older version because your ancient card has been dropped. My laptop, a Compaq Contura 4/25c falls into this category. It has this weird _QVGA_ video which AFAIK is 3.3.6 only.

Somehow though, Debian has managed to port the 3.3.6 XF86_SVGA xserver to 4.1.x, so I could potentially install the latest version. I did this for my friend, he has Cirrus Laptop Mystery Video which worked with 3.3.6 but not 4.1.x, the Debian backport fixed him right up.

X is really a fantastically stable platform. It is great that the X team is working away, but don't feel like you _have_ to upgrade just because a new version is out. The new versions are mainly made to support new hardware. If your hardware works ok then you do not necessarily need to upgrade unless you just want to.

Re:Nvidia Drivers?

[Anonymous Coward]

AGP does work on nForce chipsets. You can't use agpgart, since it doesn't support nForce. You need to use nForce's video drivers' nvAGP option.