Slashdot Mirror
XFree86 4.0.1 Review
Fawking DSL writes: "BSD Today reviews the new features, supported cards, and installation of XFree86 version 4.0.1." The article starts off by saying that XF864.0 "Shocked The World" which I find mind numbingly amusing, but it's a good review.
Antialiasing of fonts was invented because screens had fairly low resolution and so fonts looked jagged compared to typefaces in the real world, like books and magazines.
No, it was invented for images to look better. Antialising is not used just for fonts. Take a look at a web site with someone who didn't antialias their images, and it looks awful even at high resolutions. Granted, at a high enough resolution you wouldn't be able to tell, but we're talking about thousands of pixels per inch before you wouldn't be able to tell.
Also, you seem to be under the impression that photographs are not "antialiased", but they are (although, it's not called that). One of the ways you can tell a fake photograph is that the edges of an image are too clean.
it just makes the text look more hazy and less well defined, which puts more strain on the eyes.
I know that some people feel this way, but you're by far in the minority.
--
Sometimes it's best to just let stupid people be stupid.
I would recommend like this
There is much cruelty in the universe, John.
Yeah, we seem to have the tour map.
Anyway, if you are looking for the mysterious 0.94 drivers for the NVidia chipset, I found them on the bot on irc.openprojects.net #nvidia.
My bot (iCE-DCC) also has files you simply cannot get anywhere else, including usually many inter-release "test" or "experimental" drivers which fix some bugs, and right now gcc pentium pro+ optimized 0.9-4.1 drivers. Again you can check it out in #nvidia on irc.openprojects.net
-- iCEBaLM
While it would be a Good Thing© for the end user to have NVIDIA to open their specs, it would very much be a Bad Thing© for them as a company.
IIRC, NVIDIA has licenses that do not allow them to open up all the specs of their hardware. I'm sure they would love to open everything up if they could -- they could shift development efforts out of the house and score PR points with the Linux community. A while back, /. linked to a set of gaming benchmarks showing NVIDIA was well ahead of the pack in terms of X/OpenGL performance, including 3dfx which claims to be supportive of OSS.
I think NVIDIA is committed to supporting OSS, but is simply unable to practice it themselves.
I registered my hate for Jon Katz
Did you actually *look* at the screen for this comparison?
I run my 17" monitor in 1024x768; anything more, and you're begging for eyestrain. (I know, because I try to run 14-15" monitors at 1024x768, and that can get somewhat tiny... 800x600 is a reasonable compromise)
Netscape's fonts look *horrible* under X, as do any word processing program you'd care to mention under Linux that doesn't do it's own anti-aliasing.
Anti-aliasing is great for making anything look smoother. If the object in question is too small, or if you try to do it too much, yes, it might end up looking blurry. But that's a lot better than having it look, say, deformed or unreadable, like the output of a fax machine.
In Windows, True Type fonts are very scalable, and tend to look better. But they also benefit from anti-aliasing, which is a built-in Windows feature. Hey, if you don't want it or need it, don't use it. But I'm telling you that fonts in X often look horrible without it. So, check it out for yourself before you spout off again. Please.
---
pb Reply or e-mail; don't vaguely moderate.
pb Reply or e-mail; don't vaguely moderate.
While it would be a Good Thing© for the end user to have NVIDIA to open their specs, it would very much be a Bad Thing© for them as a company.
I love how you say that with such certainty, especially since you have no knowledge of any "secrets" they're hiding. Opening specs is ALWAYS benificial to a company, as that makes their hardware usable on more systems, as third party coders can write the drivers, which cost the company absolutely nothing.
IIRC, NVIDIA has licenses that do not allow them to open up all the specs of their hardware. I'm sure they would love to open everything up if they could -- they could shift development efforts out of the house and score PR points with the Linux community.
Yeah, except they wont say who they have these phantom licenses with. The only two players they do name are SGI and VA Linux, both of which are extremely Pro-OSS and would have absolutely no problem letting nVidia open their code. The truth of the matter is nVidia is an old school company who is absolutely paranoid. They're in the lead right now and they want to keep it that way as long as possible, even giving a hint of how their hardware works fightens them beyond belief. "Oh but out competators could use that information to compete with us!" Yeah right.
They're using the excuse of phantom third parties having licenses with them which prevent them from opening up specs, and in reality, they don't exist.
I think NVIDIA is committed to supporting OSS, but is simply unable to practice it themselves.
Think again...
-- iCEBaLM
For NVidia, although, this support didn't come easily. Several days after XFree86 4.0.1's release, reports of problems all over starting springing up-and all fingers were pointed at NVidia. It seemed that NVidia's 0.9-3 drivers were not up to par with the new XFree release and caused much commotion between NVidia users all over. A week and two patches later, NVidia support is finally up and going and a new series of drivers, 0.9-4 have been released, although not publicly on NVidia's website.
The 0.93 drivers not being up to par is an understatement. If you have a TNT2 card you can expect the X server to segfault on startup. After installing the 0.94 drivers it started right up for my but I still occasionally get some corruption on the desktop--other than that they seem to work nicely.
Anyway, if you are looking for the mysterious 0.94 drivers for the NVidia chipset, I found them on the bot on irc.openprojects.net #nvidia. Supposedly they're on fileplanet also, but I couldn't locate them there. Just figured I'd save the TNT2 users the hassle of waiting for NVidia to put them up on their site.
numb
But what about us non-linux users?
XFree86 4.0.1 introduced that neat binary module
thingy, and here comes nvidia releasing part of
their driver as a linux kernel module!
I noticed this AFTER I ordered a TNT2... I was
like, hey, it works with XFree86 4.0.1, I'll
buy the thing... I'm running FreeBSD... is
there any hope? Grrr.
Right her e.
The NVIDIA Linux FAQ has been updated too. In particular, 0.94 Changes
Excuse me, but NVidia released the X 3.3.x driverrs as full open source including GL drivers. IT's just their X4 drivers that are closed. If you want the source, go work with the 3.3.x source which is still available on their website. Start a CVS. Do whatever the fuck you want. But at least think about what NVidia has done before you spout off about OSS.
Ahh yes, the nicely obfuscated and useless 3.3.x drivers, how useful. Those drivers were slow as hell, didn't use any advanced fuctions of the cards, and are generally useless, so why even bother with them?
Sure, we don't know why X4 drivers arent OSS yet. Don't worry about it. You probably don't even on an Nvidia card, much less know enough about it to think that you could do a better job programming drivers for it than its creators could.
Yet? You're saying yet as if there is any hope that they WILL be in the future. I'm telling you there isn't. The only part of the drivers which has any hope in hell of being open is the XF86 4 XAA driver module, the GL/GLX implimentation and the kernel driver (except for the kernel interface) will always be closed. And if you want to talk to me about how my TNT doesn't exist you're quite welcome to come by #nvidia on irc.openprojects.net and say that to me with a straight face.
Yes, I do belive the community could write better drivers then nVidia for the simple fact that these drivers share too much code with the windows ones, and nVidia has almost zero experience coding for Linux. Hell the windows drivers don't even honor monitor definition file set modes! I have to use a third party program to switch the sync polarities at run time to even use windows.
Nvidia is not an old-school company. Microsoft is an old-school company. Nvidia is making drivers for Linux, aren't they? Closed source is not old-school, otherwise few people would be still doing things that way.
nVidia is an old school company, their suits follow the old school "keep it to yourself" system of information handleing.
Nvidia does not need to open up precise specs just so that their card can be used on Linux. They wouldn't have had to do that for any other OS. The linux community, in "requiring" companies to OSS their drivers and programs, is effectifly turning away potential suppliers.
No, they don't have to open up specs so the cards can be used in linux, they can just let 2 or 3 developers code half-assed drivers which lockup and crash when you look funny at them. Do you know that switching to a virtual console while in X with the current drivers will lock up your system? Do you know that for many monitors you cannot specify modelines because the drivers think they know best by autoprobing? Do you know that there is even a report that running "tar" without parameters in X while using the latest drivers will result in a system reboot? These drivers suck, not because the developers are incompetant, but because there arent enough of them and they're inexperienced when it comes to writing for X.
Besides, when there are only 5 or so 3D games for Linux, why bother worrying about it? Windows has game dominance for the forseeable future.
What an obvious troll. There are many more than 5, MANY more. And it's not just about games, its about blender, xmms plugins, q3radiant, and other little niceities you'd like to use because you bought this super duper 3D card. Not to mention Linux gaming wont get any better if there aren't viable 3D accellerator options available.
(I have a TNT2 and dual boot between Debian and 98 on a regular basis. So I know to a good extent what I am talking about.)
I really don't think you do.
-- iCEBaLM
X will be around for a while to come, and there are a few reasons I'll outline below.
Be realistic. So the x86 port of OS X comes out (No, it probably won't). Why won't it? Because Apple have too much to lose in terms of hardware sales, if people could run OS X on a $1000 PC instead of a $2000 Mac, would they really choose the Mac?Secondly, X isn't tied to one platform. (By platform I mean hardware and software platform). It runs on Alphas, Sparcs, SGI boxes, Intel, PPCs, you name it. And on those platforms, (hardware), it runs on several software platforms running on those hardware platforms. So an OS X port for Intel comes out. Even if all the Intel UNIX users switced over to OS X, X Window System would still be more than alive and kicking on dozens of other platforms.
In other words, we should be commending the authors of this new implementation of X11, because without it, UNIX would have no standard GUI. Like it or not, (and you should, because it areas that it really counts, X11 is way ahead of other systems, ever tried multiple remote sessions on a Win 98 or NT box? Oh, sure....let's buy the Citrix WinFrame software or Microsoft's new ultra-expensive Terminal Server, both of which are overpriced and don't perform as well as stock X). X is here to stay for quite a while to come, and improvements are welcome. Congratulations, XFree86 Team!"A few atoms won't even light a match" - Dr Jones, 1933
I suggest the whole Alien series, Alien I, Alien II, Alien III, Alien IV, and maybe by the time the compile is over they'll release Alien V.
When closed-source/specs video manufactors have gotten around to openning up the programming specs for their cards, there usually isn't any "trade secrets" that wheren't already known by the video card industry. It is more often that openning up the specs ends up revealing the bugs and needed work-arounds with the video card's implimentation. So, I think this begs the question, if Nvidia thinks they are such hot stuff, why do they feel the need to continue to cover up their flaws? For Nvidia to have been over backwords to continue to keep their latest video cards only usuable via closed-source drivers seems to indicate me that the flaws in their latest stuff must be so embarassing that they feel that can't even speak publically about it. But what exactly is this massive flaw that got them scared of the peer-review that commonly takes place in the open source community? I think I have an idea. After taking a look at Nvidia and ATI side-by-side on the same monitor, I have decided that Nvidia is cutting corners on doing true-color to accomplish faster speeds. While the Nvidia, when supposibly in 24 bit color mode, does produce more colors than in high color mode (16 bit), it doesn't appear to be the full color depth produced by an ATI in true color mode. Of course, I can't test this theory by examining the drivers to see if some color bits are stripped in the driver because Nvidia is continuing to keep that covered in the cloud of closed-source. But until Nvidia is willing to accept peer-review to back up their claim of supporting true color, I'm going to continue to recommend ATI to clients.
Okay, here's a point of view that's been rarely expressed here, but here goes: font antialiasing is a crufty outdated process which isn't really very useful on today's monitors.
Antialiasing of fonts was invented because screens had fairly low resolution and so fonts looked jagged compared to typefaces in the real world, like books and magazines. Also, unlike black typeface on white paper, computer programs and Web pages used some interesting background and font colors, which could be visually jarring in their contrast. Solution: Blend the edges of fonts into the surrounding background color, and the fonts were more pleasing to the eye both because jaggies got eliminated and because the color gradient made the color transition between font and background less jarring.
Fast forward to today, and 15 inch monitors capable of 1024 by 768 are the minimal norm, and 17 inch monitors capable of 1600 by 1200 are fast becoming commonplace. So, the problem with jaggies is no longer a problem at all. The problem with visually jarring combinations of color is no longer a problem, either, because people are as used to Web and application colors as they are to standard black-on-white printing.
Font antialiasing becomes useless unless you're going to be running at very low resolutions. Otherwise, it just makes the text look more hazy and less well defined, which puts more strain on the eyes. It also impacts system performance: I noticed a significant speed boost in Explorer responsiveness when turning anti-aliasing off in Win98 on a K6-2 400. Anti-aliasing is great for games and graphics displayed at fairly low resolutions, but on the modern desktop of a contemporary OS it is unneeded and impacts system performance negatively.
"The more corrupt the state, the more numerous the laws."--Tacitus, *The Annals*
Nope, still won't work quiet right. The Rawhide RPM's you're trying just won't install on anything but Rawhide, as far as i can tell.
;)
You've got as far as the RPM version error message. I got as far as actually trying the Rawhide rpm binary with the Xfree4.0.1 RPM's...and there are a load of failed depencies that not even --nodeps will fix.
I've also tried the straight XFree4.0.1 binary tarballs with their installer, and it will, sortof, work. The biggest problem i had was that xdm didn't work with the new version of X, which meant i lost my graphical login. Which sucked.
But provided you don't use xdm (The graphical login), you should be able to uninstall the XFree3.3.x RPM's and then install the XFree4.0.1 binaries from a tarball.
Or, do as i plan to do, and change to a decent distro
Syllable : It's an Operating System
Steps to upgrade XFree86 from rawhide
This worked for XFree86 4.0, I'm on step 4 right now :). Remember, the rawhide packages are sometimes broken, if it doesnt work, try again in a few days.
treke
Being a big Glide user, i was disappointed at the fact that you can't use Glide with 4.0.1 (You can, but it's buggy as shit! usually crashes after one glide app).
Does anyone know if there is anything in the works to get stable glide support in addition to DRI in the new releases?
FluX
After 16 years, MTV has finally completed its deevolution into the shiny things network
"It is seldom that liberty of any kind is lost all at once." -David Hume
I have a g400, and I really like it, it's 2d signal quality is far better than my riva tnt based asus card. The geforce is probably much better in 3d, but the 3d of the g400 is rather good, and it has dual head on a single card. It works well with linux, freebsd, openbsd, beos 5 etc. But I can't get it to work with plan 9...
I am trying to set up the g400 with xfree 4 dual head, but I haven't succeeded, and I can't find any clear documentation on that (any tips anyone?). Single head works well with xfree 4 with mandrake linux.
NVIDIA have publically stated that they cannot open their source because they use a third party AGP driver implementation which they do not have a license to release to the public. This code is responsible for a great deal of their performance.
HAHAHAH! Thats the funniest thing I've ever heard! God. Show me where they state this because if this is the only reason they're keeping them closed then they could open the drivers tomorrow.
1. Their "third party AGP implimentation", if this is really true, and I highly doubt it, doesn't even WORK on most AGP chipsets! It only really works on intel chipsets. So NVidia is getting ripped off.
2. The open part of the kernel module was modified by an open source coder to add support for kernel AGPGART instead of using the internal NVidia AGP routines. It isn't as fast on chipsets which the NVidia AGP implimentation works with, but god, it works on a LOT more chipsets, and its not that slow, maybe about 5 FPS loss in Q3, big deal.
They could spend lots of time and money to get rid of this code, just for the sake of being Open Source, but it doesn't make business sense at this time since they already have a kick ass driver. Not to mention that they would probably get immediately sued by the AGP driver company the instant they "came up with a new implementation" and stopped paying royalties or whatever.
Hah, like I said AGPGART support is already there, clear the old one out and switch to that, the actual implimentation is in the kernel, wouldn't have to put it in the driver itself, would cost almost 0 dollars, and there is no chance of suing. Kick ass driver, rofl. Have you even used them?
-- iCEBaLM
Probably not as much as our new "governor"...
-JD (A Minnesotan)
- Install RPM 4.0.
- Download the RPM 4.0 tarball ( ftp://ftp.rpm.org/pub/rpm/test/rpm-4. 0.tar.gz ).
- Download db3, the database format used by RPM 4.0 ( ftp://ftp.rpm.org/pub/rpm/ test/db3-3.1.14-0.2.6x.src.rpm ).
- Build ( rpm --rebuild db3-3.1.14-0.2.6x.src.rpm ) and install ( ls
/usr/src/redhat/RPMS/i386/db3-* | xargs rpm -ihv ) db3. Build ( rpm -tb rpm-4.0.tar.gz ) and install ( ls /usr/src/redhat/RPMS/i386/RedHat/RPMS/i386/rpm-* | xargs rpm -ihv ) RPM 4.0. - Convert the RPM database to the format used by RPM 4.0 ( rpm --rebuilddb ).
- Moving the old database files (
/var/lib/rpm/*.rpm ) out of /var/lib/rpm is necessary to prevent RPM 4.0 from segfaulting.
- Update initscripts, modutils and chkconfig to be compatible with those expected by the rawhide distribution of XFree86 4.0.1.
- Download the updated packages.
- initscripts 5.27 ( ftp://download.sourceforge.net/pub/mirrors/redhat
/ rawhide/SRPMS/SRPMS/initsc ripts-5.27-1.src.rpm ) - modutils 2.3.11 ( ftp://download.sourceforge.net/pub/mirrors/redhat
/ rawhide/SRPMS/SRPMS/modutil s-2.3.11-7.src.rpm ) - chkconfig 1.2.1 ( ftp://download.sourceforge.net/pub/mirrors/redhat
/ rawhide/SRPMS/SRPMS/chkconf ig-1.2.1-1.src.rpm ) - Build and install the updated packages.
- rpm --rebuild initscripts-5.27-1.src.rpm ; rpm -ihv
/usr/src/redhat/RPMS/i386/initscripts-5.27-1.i386. rpm - rpm --rebuild modutils-2.3.11-7.src.rpm ; rpm -ihv
/usr/src/redhat/RPMS/i386/modutils-2.3.11-7.i386.r pm - rpm --rebuild chkconfig-1.2.1-1.src.rpm ; rpm -ihv
/usr/src/redhat/RPMS/i386/chkconfig-1.2.1-1.i386.r pm
- Build and install XFree86 4.0.1.
- Download XFree86 4.0.1 ( ftp://download.sourceforge.net/pub/mirrors/redhat
/ rawhide/SRPMS/SRPMS/XFree8 6-4.0.1-0.30.src.rpm ). - Build XFree86 4.0.1 ( CFLAGS='-I/usr/src/redhat/BUILD/XFree86-4.0.1/xc/
e xport/include' LDFLAGS='-I/usr/src/redhat/BUILD/XFree86-4.0.1/exp ort/lib' rpm --rebuild XFree86-4.0.1-0.30.src.rpm ) and install the desired XFree86 4.0.1 RPMs located in /usr/src/redhat/RPMS/i386/XFree86-*. Set CFLAGS and LDFLAGS is needed else the build will complain about being inable to find some parts of xlib, which should only an issue when building on a system without XFree86 installed.
I haven't really had any problems except for warnings in the USB portion of rc.sysinit, which can be commented out. Some files were renamed or were moved around as well (i.e., conf.modules becomes modules.conf and the init scripts are moved fromNumlock key which actually works. Like it was in 3.3.
Actually, it is supported. I have the dual-head version too, and it's not bad at all.
"If ignorance is bliss, may I never be happy.
-- Veni, vidi, dormivi