Hardware Manufacturers that Actively Support Linux?

wirefarm asks: "I know there is are lot of well-supported pieces of hardware for Linux, but I was wondering, which vendors really go out of their way for the community? While tracking down drivers for a wireless PCMCIA card today, I found that the vendor boasted of having Linux support, but it was seemed that they were actually touting drivers that were community-developed, rather than written with any help of the company. So my question is this: Which companies really stand out when it comes to providing specs and developing drivers?"

Comments

Not yet!

[samjam]

Canon don't yet, I was very annoued with my facncy new cheap 650 USB scanner!

They are still "thinking about it" and won't give out any specs in the meantime.

Re:Not yet!

[volsung]

Note to people looking for USB scanners: Epson has apparently been nice to the developers. From the SANE USB scanner list:

Epson have[sic] been very helpful with the development of the backend, to the point of providing documentation that's not yet released.

Re:Not yet!

[rafelbev]

This also applies to their printers. They are the most supported... much more than HP does. Epson printers simply ROCK regarding linux support. My Stylus never gave me problems and I know alot of people who can say the same. They guys who import them in my country support our LUG too.

Just my 2c

Rej

Re:Not yet!

[prizog]

"Epson have[sic] ..."

No, not sic. It's a British (and Australian) convention to treat a company as a group of individuals, a plural. This makes a lot more sense than American, which can't make up its mind whether companies are singular or plural. Both of the following are acceptable in American, although the first more so: "IBM is the leader in memory technology; they have just released a new 1TB memory module." "IBM is the leader in memory technology; it has just released a new 1TB memory module."

Disclaimer: my Australian sample size is 1, and my British sample size not much larger. I'm an American who is trying to switch to the British convention for obscure political reasons (I don't like the idea of companies as entities comparable to individuals -- it removes responsibility and encourages unethical behavior).

Re:Not yet!

[pmz]

(I don't like the idea of companies as entities comparable to individuals -- it removes responsibility and encourages unethical behavior)

In the U.S., my incentive to incorporate my business is to insulate myself and my assets from my business and its assets. This is an important legal protection. I'm not really convinced it encourages unethical behavior; rather, it encourages people to take on greater risks, which are very important in growing innovative businesses.

Some people may try to use this mechanism to mask accountability, but it really doesn't work in the long run. Healthy businesses have to maintain accountability to stay strong, since a business is fundamentally a group of people working together.

Re:Not yet!

[Zaknafein500]

Because in English we have 3 options for sex, male, female, and unspecified. It's not like Spanish where everything takes on a gender. A company is obviously neither male, nor female, so you refer to it as "it".

Re:Not yet!

[Drachemorder]

Interesting. I haven't been able to get my Z42 working in Linux yet, although I admittedly haven't tried very hard.

Re:Not yet!

[LinuxGeek8]

Besides Epson, I believe Umax did release the specs of their scanners also.
And Mustek (or was it Plustek?) developed some drivers for their scanners.

Re:Not yet!

[LinuxGeek8]

Maybe Umax released just the specs for the scsi models.
I am not sure about that.
But yes, support for parport and usb Umax's are rather new.

Re:Not yet!

[big_hairy_mama]

It sounds like your printer works fine, but Lexmark has *not* been kind to Linux with its other models. Fortunately, my model (the 3200) was reverse engineered, but certainly not with Lexmark's support. Some other models have also been so fortunate as to be reverse engineered by some intelligent Linux developer, but Lexmark in general will give you nothing but hell. The 3200 driver isn't even maintained anymore, so good luck with that.

My Lexmark has been a good and reliable printer, but I would *not* recommend it to someone looking for a Linux printer. Go with Epson or HP, as earlier posts have pointed out. Better yet, get a Postscript printer like the old days :)

Intel

[swagr]

My Compaq Evo n600c laptop had an eepro100 that wasn't supported by the kernel until 2.4.18.

Intel had a src download driver that compiled and worked flawlessly.

Re:Intel

[swagr]

Yes BUT, the kernel would fail to recognize that it could support my particular card, and hence would not. This made the kernel's "support" completely useless in my case.

Re:Intel

[DragonWyatt]

I just got it working.

Check here to see how.

Re:Intel

[DragonWyatt]

For commentary (and a sort of How-To) on my interacting with both Compaq *and* Intel on trying to get the multiport wireless option to work under Linux, check here.

Summary: Neither Intel or Compaq could/would help me, even discouraged me by telling me it couldn't be done. But, with a nice set of tools from these guys, I got it to work, and was even able to contribute to the project so that this device would be supported in the future.

Oh yeah. I also CC'ed Compaq and Intel's technical support ;) .

By the way, the builtin eepro100 was supported well by rh7.2's stock kernel, FYI (proof that integrators and distro maintainers are valuable methinks).

Re:Intel

[Dwonis]

Do an lspci -vv (or something like that) and send the results to the driver's maintainer. It's probably a 1- or 2-line patch.

Re:Intel

[swagr]

Your comment is very insightful. Thanks. I'd mod you up if I could.

nvidia, but...

[Dead_Smiley]

they are not Open Source. I guess this is the obvious one to many... mode me down if you wish.

Re:nvidia, but...

[knewman_1971]

Call me crazy, or mod me as flamebait, but...

Frankly, I couldn't care less if nvidia's drivers are open sourced. After spending months trying to play Quake II on a Voodoo5 5500, I bought a GeForceII MX 400. I was playing within 5 minutes of installing the card.

I've owned an Intel Pocket Concert MP3 player for over a year...still can't use it on Linux...(yes, there is a project in ALPHA on freshmeat...and it's been in Aplha for the same ammount of time that I've owned the player.

My concern with Linux drivers for hardware begins with "If the fscking thing supported at all?" and ends with "Hmmm. WHich kernel am I going to have to use today?". If a vendor actuallly takes the time to give me drivers, then fantastic. I'm just not going to quibble about the open source thing.

I'll fight that battle when MOST vendors include drivers. Until then, I'm happy just to be able to use my shiny toys.

Re:nvidia, but...

[fizz-beyond]

Frankly I'm with you, there are alot of people here on /. that would stone you if they were given the chance because you didn't release something opensource, and quite frankly I think it's too extreme.

The choice of what license to use must be made completely based on the project. I assume in nVidia's case they don't want to give up the specs because they feel that it would help enable people to reverse engeneer their product (that's only a guess), but they still want to support free software.

Re:nvidia, but...

[(startx)]

IMHO, nividia is awesome. I don't care who has issues with the license of their drivers, at least they work. Not only that, but I get higher framerates in slackware than I do in windoze 98. As long as the nvidia binaries remain rock solid and work out of the box, I really don't care if I have the source or not.

side rant: ATI on the other hand releases the specs, but seems to do no actual work themselves. This does help produce free drivers, but they take forever! My friends radeon 8500 STILL doesn't work in XFree fully, while my gf4 ti 4600 has been humming along nicely since the day I bought it.

Re:nvidia, but...

[defile]

Binary only drivers are inferior. Even when you have an open sourced kernel module to intermediate. The argument would be less unreasonable if it was source vs. open source, but it's not. It's binary only vs source available.

In any case, nVidia wants to open source their drivers. The reason I got for them being binary only was that they licensed the AGP code from a third party which is unwilling to open their code. Too bad.

Re:nvidia, but...

[vidarh]

No they have not. Try actually looking in the archives next time.

Re:nvidia, but...

[Eric Smith]

nVidia wants to open source their drivers. The reason I got for them being binary only was that they licensed the AGP code from a third party which is unwilling to open their code.

If that was really true (which I doubt), nVidia could release the sources with the third-party code stripped, and the community would write a replacement for that part.

The reality appears to be that they think by releasing sources or programming specs, they'll somehow make it easy for a competitor to clone their chips. But as any ASIC engineer knows, that's not true. If it were, everyone would be making Pentium IV clones, since the specs for that are published. The reality is that designing a chip with tens of millions of transistors is a very large amount of work, even with the programming (register) specs.

nVidia did release some source code at one point, but it had been run through the C preprocessor, so it was effectively obfuscated.

I used to buy nVidia-based cards, but now I prefer ATI or Matrox. They may not be as high performance, but to me the support is much more important. Anyhow, I have yet to find anything I do for which the performance of the ATI or Matrox cards is inadequate. I don't have any need for frame rates above 72 Hz.

Re:nvidia, but...

[Ogerman]

Frankly, I couldn't care less if nvidia's drivers are open sourced. After spending months trying to play Quake II on a Voodoo5 5500, I bought a GeForceII MX 400. I was playing within 5 minutes of installing the card.

If you're a gamer junkie, then fine. For those of us whose computers are more than just toys, NVidia's unstable closed drivers are not even a consideration. And maybe you'll reconsider as well someday when X crashes while running a GL accelerated app. Closed source drivers are 100% illogical and unacceptable under any circumstance. That's all there is to it.

Re:nvidia, but...

[knewman_1971]

Perhaps I should have thrown in more hardware horror stories...

* Digital Capture Device (w/USB)
* My USB scanner.
* Any of my inkjet printers.
* The freakin' southbridge on my mobo.
* Every webcam I've ever tried, with the exception of my Creative Labs cam


...trust me, the list goes on. My computer is a toy, from time to time. But most of the time, it's coding, designing, capturing, or something less-than-entertaining.

I'll concede that if nVidia drivers caused my system to crash, I'd be the first to chunk the card and drivers, and go looking for an alternative. But, I've run Mandrake, RedHat, and Suse (various versions) on the four cards I've owned, and I can tell you that MY only complaint is the splash screen. Which can be turned off.

But, back at the point, even if I went shopping for a new video card, my argument remains the same. I'm not going to quibble about the source code. If it's open and available, great. If not, and it somehow manages to work despite that, great.

From my perspective, as an end user, even if the source WAS available, I couldn't do a damn thing with it anyway. Nor do I know anyone who could. Sure, the "community" could probably pull a rabbit out of it's collective ass and build one. But how long do I have to wait? It's not my fault I'm not a device driver programmer.

Don't get me wrong. At the end of the day, I think open source is a great idea. I'd love for all drivers and software to come with clean, well documented code. But it doesn't. I'd also love to have a group of highly skilled developers on staff who did nothing but write device drivers for hardware that I own, or would like to own. But again, I don't.

So what's the alternative? Do I have to accept the fact that wanting to run linux precludes me from running the latest hardware?

Sadly, the answer is yes, in lots of cases.

For me, the answer is clear. I'm going to give as much business as possible to vendors who are kind enough to throw a bone our way occasionally, as opposed to beating them up for not coming all the way. More power to you if you choose to do it another way.

Re:nvidia, but...

[Ogerman]

Maybe this is because the Linux kernel device driver framework is so poorly done and un-standardized than you NEED the driver source code to do anything.

Maybe you don't have a clue what you're talking about and have never even looked at the kernel source. Some drivers are better written than others, but the framework is remarkably solid.

ATI

[CynicTheHedgehog]

ATI gets a lot of bad press for their drivers, but they do release the specifications for their hardware to multiple open source development groups. What you end up with is Free, open drivers that are as good as the groups that make them. This as opposed to NVidia, a company that although support Linux through binary drivers, does not release the source code or specifications.

Re:ATI

[Sabalon]

Oddly enough, as you were trying to say that ATI is a little better than Nvidia (cause of the opensource/binary driver thing), you have said that NVidia activey supports linux - goes out of there way as the question said, while ATI just says "here...good luck".

ATI gets a lot of bad press for their drivers for a damn good reason too :)

Re:ATI

[CynicTheHedgehog]

while ATI just says "here...good luck"

Circa 1998, this was all anyone ever wanted. Remember the OSS (sound for linux) project? They claimed that if someone bought them a board, or gave them the specs, then they would write a driver for it. And they did, too. I suppose it's reasonable to expect a company to produce drivers for Linux, but remember, there are umpteen billion operating systems out there, and these companies don't have the time or resources to develop for all of them.

Personally, I'd rather have the specs and free drivers that anyone could hack on. I'll bet the NVidia/AMD issue wouldn't have lasted a week (hell maybe not even a day), and with time people will hopefully no longer have a reason to bitch about drivers for ATI hardware.

Re:ATI

[EasyTarget]

... Of course, there is an argument that just giving the hardware specs and saying 'good luck' is the open source route..

Since if RMS etc are to be believed hundreds of us(*) then jump in, write free (beer/speach) drivers, if something does not work it gets tweaked/fixed etc..

[Of course that is a bit trollish] What shows a true commitment to the open-source customers is community development, with the manufacturer releasing HW specs, but with them also making some technical resources available to answer the really difficult questions driver development often poses (awkward timings/settings/etc..)

(*) Well.. not me obviously, I can't program C/CPP/ASM at all, and I guess nobody wants drivers written in Perl ;)

Re:ATI

[johnjones]

considering that they EMPLOY people to work on XFREE86 I dont think that they should do anything else

what should they do ??

oh and the ATI MIPS SOC chip has a linux port

so really ALL of their chips graphics/CPU's support linux and thats better than 99% of the others and definately better than NVidia closed source fscking stuff

regards

john jones

Re:ATI

[CynicTheHedgehog]

I remember reading somewhere that ATI developers are allowed and I think encouraged to participate in Open Source projects. This is hearsay of course, but it would be nice if it was true.

Re:ATI

[lkaos]

FUD, FUD, FUD, FUD *to the SPAM theme*

ATI only releases a portion of the specs to a small group of developers who have signed an NDA.

ATI does not release specs for features that are unique to ATIs cards.

So for stuff like hardware iDCT and TV-OUT, ATI has released absolutely no specifications to anyone.

Not to mention the fact that they do not like to give their specs out much so you end up with like 3 people who actually have the specs and everyone else has to reverse engineer the drivers to figure anything out.

Re:ATI

[Darren Winsper]

Yeah, the drivers for the Radeon 8500 were flawless at launch, not.

When the Radeon 8500 was launched, the Windows drivers didn't even support Smoothvision, which was a major selling point of the R8500. Oh, let's not forget the Quack3 fiasco, or the poor performance, or the poor stability.

Still, it was better than the XFree drivers at launch, there weren't any. Whereas with the NVIDIA drivers, there are typically official Linux drivers before the card is available.

Re:ATI

[Performer Guy]

NVIDIA actively support Linux by constantly releasing up to date drivers that are very high quality. The NVIDIA drivers are unquestionably the highest quality OpenGL implementation available on Linux without exception. ATI supply the specs but apart from that do almost nothing, they have enough trouble supporting high quality Windows drivers. The reason you need the specs is to get any kind of driver support at all, when the manufacturer is delivering full high quality up to date drivers with more OpenGL support and extension support and quality than anyone else I'd rather have that than specs and a driver development effort that can't keep up. OpenGL is not like most other driver efforts, there is a level of complexity and testing required which seems to require more support and maintenance and a higher level of expertise to get high quality than is currently applied to them by the Open Source community. I'm disappointed by people who constantly feel the need to dis NVIDIA when they do more to support Linux than any other hardware vendor, simply because the way they choose to support Linux doesn't match your philosophy.

This is not a vendor who ignores Linux, they give Linux fantastic support at a level beyond any other hardware manufacturer due to the complexity of their effort. It also produces better results than the driver development models you espouse.

Re:ATI

[Darren Winsper]

I'm getting sick of this "SGI contract" thing. There have been SGI people posting on Slashdot who point out its false, yet people still proceed to claim this.

I tell you what, I'll shut up if you can point me to one authoritive source of this claim.

Re:ATI

[psamuels]

Actually, you can get the sourc for the nVidia drivers for the purpose of compiling it for your own machine. You're just not supposed to read it.

Have you looked closely? Last I remember, you got the source for a small part of the drivers for the purpose of compiling on your own machine. The deep magic still happened in the *.o file(s) you got no source for. The driver source you got was just a shim layer to make sure the Real Driver could work correctly with your current Linux kernel.

Has this changed?

Re:ATI

[max cohen]

I'll bet the NVidia/AMD issue wouldn't have lasted a week (hell maybe not even a day)

IIRC, the Athlon/AGP issue isn't just related to Nvidia products, it happens on ATI products as well. Since Nvidia wrote their own driver, they may have made some error which causes the problem to occur more often.

Re:ATI

[Performer Guy]

This is a lie. NVIDIA is not prevented from disclosing driver source because of contracts with SGI. Heck SGI even open sourced their OpenGL SI a while back, that's basically their reference driver. I agree that NVIDIA has great Linux drivers and applaud them for it (see my other post) but their reluctance to release source code is not because of SGI.

Perhaps they think a company open sourcing drivers and microcode increases their exposure to patent infringement law suits, and IP theft. Or perhaps they have issues because they have agreements with M$ because their code base is contaminated by their SDK (doubtfull).

Re:ATI

[larien]

Hrm, I remember playing linux Quake III in a train using a laptop with an ATI Rage Lt Pro; isn't that the same chipset?

Re:ATI

[Fluffy the Cat]

NVidia activey supports linux

No, NVidia actively support Linux/x86. Want to use a GeForce in an Alpha? Oops. By releasing documentation, ATI allow their hardware to be used on all Linux platforms rather than a subset of the popular ones.

Re:ATI

[Performer Guy]

You don't understand the complexity of developing high quality functional OpenGL drivers. Each card requires a major project if you are going to do more than support a core subset of what the hardware is really capable of.

Re:ATI

[bokeoa]

A lot of good that does for people trying to get an NVidia card working on a Mac.

Re:ATI

[Hamshrew]

As another person pointed out, that does little good when trying to use the Nvidia cards on another platform. While the binary driver is their choice, and I applaud the work they have done, there are other reasons to choose an open-source driver.

As for ATI doing "almost nothing," they were, until very recently, paying developers to work on their open-source drivers, in addition to releasing specs, which was all the community asked for.

Re:ATI

[dinivin]

Or FreeBSD.

Or heaven forbid you want to run the latest development kernel.
And don't even think about trying to run two nVidia cards at the same time with their driver. In fact, I couldn't even get my nVidia card to play nice with a PCI Permedia 2 card.

Frankly, I'd rather not put up with crap like that :-)

Dinivin

Re:ATI

[CynicTheHedgehog]

It wasn't my intention to criticize NVidia; I was trying to point out that both companies support Linux actively in different forms. Call it compare and contrast.

But, to be perfectly fair NVidia has had a jump on ATI as far as Linux drivers and X goes. The GeForce series has been around a lot longer than the Radeon. It would be nice if ATI was more active and open as far as the development process is concerned, but I have faith in the people who are currently working on it.

Re:ATI

[CynicTheHedgehog]

To my knowledge the ATI drivers are included with XFree86, and support for other features (TV in, etc) can be achieved by downloading source or binary packages for other applications (since neither the kernel nor X can really do that stuff). I believe you have to download NVidia driver binaries from their website and agree to some sort of EULA. It's six of one and half a dozen of the other. Someone somewhere is always going to take a source package and make it easy to install, whether it's a distribution maintainer (RedHat) or some industrious software developer.

Re:ATI

[Surak]

While it is certainly commendable that nVIDIA constantly releases new drivers for their cards on Linux, I'll take specs over closed source drivers any day of the week.

Here's why:

If you're always updating your hardware and you keep updating to the latest and greatest video cards, and you don't update your operating system too much, and you don't mess with experimental or drivers for other hardwre in your system (which means that you follow the same basic pattern as a gamer), then great, nVIDIA's development model probably works for you.

However, what if you're still using the GeForce 2 card and plan on using it for at least a couple of years, but you want to keep trying out new Linux distributions and versions, and maybe even experimental kernels? And what if nVIDIA suddenly stops developing new versions of the geForce 2 driver, because they discontinue the card and then have no funding to work on drivers for discontinued cards?

Well, as we all know, older video drivers may or may not work with new versions of XFree, new kernel versions, etc. without being recompiled. If the source isn't available for recompiling, and nVIDIA isn't doing it, who is? No one. So now you're stuck with a piece of hardware that won't take you through to the next version of the OS. You can either A) get another card, or B) not upgrade your OS.

In the Open Source model, at least you have a reasaonable assurance that if the developers of the driver drop it, someone will pick it up, even if that someone ends up being you. :)

Not everyone constantly upgrades their hardware. I know this may sound like an alien concept to the gamers out there, but its true. :)

Re:ATI

[tps12]

Has this changed?

No, this is still accurate. Major PITA, since the GLX X module needs to be the same version as the kernel module, so even though you can install different kernel modules for different versions of your kernel, you are limited to one GLX module at a time. So every time you reboot to a different kernel you need to recompile the GLX module. :(

Re:ATI

[dinivin]

Wrong... The "source" that you take about just provides for bindings between the kernel and the binary-only GLX driver. And if you're not using an nVidia blessed kernel, tough shit.

Dinivin

Re:ATI

[Etyenne]

Sure, and the day they'll decide that the 5% (or probably lower) market share they get from Linux user is'nt worth the effort they put in releasing Linux driver and stop updating their driver for Linux, what am I gonna do ?

I could continue to use the last released driver. But it probably ain't gonna work with newer kernel. I could stick to the latest supported kernel, but I suppose it's going to piss me off when I will want to upgrade to kernel 3.6.8 in a few year to get support for a feature I really need.

But then, I could upgrade my video card to one supported. I suppose that is what the hardware industry would want me to do. Unfortunately for them, I am not fond of spending money on hardware unless it is really necessary. I am still using parts I bought 6 years ago. I don't expect my video card to satisfy me that long, but I want to get at least a few years out of it.

I am shopping for a video card right now. I will not take the risk of being orphaned by Nvidia, so my money will go to ATi.

Re:ATI

[Afrosheen]

Yeah that'll take a whole 6 minutes out of your day ;P

Re:ATI

[tps12]

Well, in my experience, I'm usually rebooting trying to get some driver to work (recent example is PPP when I got my DSL...I don't like to have unused modules lying around...no, I don't have a good reason). So, lots of kernel recompiles and mucking about before I figure it out, and it ends up being irritating to have to use virtual terminals instead of xterms.

Re:ATI

[ivan256]

There are two types of stories you hear about nvidia drivers. (I speak as someone who has three machines running nvidia cards)

You see the people who have no trouble, and assume that th because they have no trouble the drivers are great, and hence nvidia is great. And then there's everybody else. Each of my 3 machines have had the X server die on occation running the nvidia driver. I have never seen XFree86 die when running any of the open modules that come with it. The module from nvidia doesn't like if you use 2 cards, wether they be both nvidia cards or otherwise. The nvidia driver doesn't always properly put monitors to sleep when it blanks the screen. I have lost a monitor to this bug. This is what you call great support? Where are the binaries for all the other platforms? Where is the support for non-X related graphics? What if I want a dual head framebuffer console?

Are you trying to tell me that you have never had XFree86 die on you with the nvidia driver? I don't believe you. You either haven't been using it for long, you reboot into windows all the time and never have a session open for very long, or you're lieing.

Here's the point. The binary nvidia drivers for linux suck at what they're intended to do (support nvidia cards on i386 linux boxes), and that doesn't even touch on all the things that they can't do because nvidia doesn't bother letting you (like using them on a mac). The open source driver is good, but it can't do 3d, and it can't support dual-headed cards, so I'm forced to have my session disappear out of under me at random once every month or so, or go out and drop a load of cash on a new, non-nvidia, dual head card. Grrr.

This has nothing to do with philosophy, that's another issue for another time. Doing more then any other manufacturer (which isn't true, unless you only count video) isn't good enough. Why is it that if you're a corporation that buys some nvidia chips, they give you the specs so you can program for them, but if you're a consumer that buys some nvidia chips, you don't get the specs, and you aren't allowed to program for them. Why the double standard. Hell, we even pay more for each chip then some company that's buying in bulk. Is it too much to ask to want to know how to use the device you've spent good money on? What good are all the features if they won't tell you how to turn them on.

NVIDIA: if you're reading this, release the dual head specs! I don't care about 3D support, just let me implement dual head in the open source driver! (And what's up with the splash screen, why do we need to wait for that?)

--

And now, off to be modded down by all the nvidia fanboys with mod points...

Re:ATI

[T3kno]

This is based on the viewpoint that the Linux community is shrinking, or that Linux users are not using nvidia cards. Both statements to me seem somewhat untrue. I think that nvidia has realized that Linux is not just a fad, but is here to stay, and right now this business model is working for them, and for me (and a lot of other nvidia/Linux users I presume). Maybe in the future they will realize that having only a closed source solution is not optimal and will realease the specs, I hope they do this, but I have no way of knowing. What I am fairly certain about though is that while nvidia is king of the hill they are going to keep support for the Linux community because it is not shrinking.

Re:ATI

[Chanc_Gorkon]

Option "NoLogo" "yes"

Poof....no mo logo! :)

RTFM dude!

Re:ATI

[tps12]

Yeah, I know. The thing is, I tend to patch my Linux kernels rather than download whole new ones, so I can't recompile, say, the new NVdriver module for Linux 2.4.10 (since I don't have a 2.4.10 source tree anymore).

The only good solution I've thought of (but yet to implement) is to use checkinstall (which I already do) to make binary packages of GLX modules, which I save. Then it's just a matter of a removepkg, installpkg, modprobe and I'm in business.

Re:ATI

[Paul Komarek]

Are the NVidia drivers really the highest quality, without question? What about FireGL, and other "professional" cards? I expect the highest-quality OpenGL implementations for linux happen on cards with more precision than any of the consumer cards offer (though the Kyro II has more internal precision than most). NVidia's binary-only drivers are probably pretty good for gaming, for those people who don't mind "tainting" their kernel.

I'll stick with ATI, who has provided information and money for linux driver development. I have a Radeon DDR AIW, Radeon DDR 64, and a Radeon 8500 (still waiting for 3D on the 8500, but it appears to be coming). I'd stick with them even if they only provided the information.

-Paul Komarek

Re:ATI

[Performer Guy]

Sigh. Please don't confuse my thoughts with yours. You're welcome to your thoughts and feel free to expound on them, just don't ascribe them to me.

Re:ATI

[Performer Guy]

Many vendors including NVIDIA have thrown some cash at developers, for example Precision Insight before VA Linux destroyed the company. SGI even gave them some cash. That's actually easier than a real commitment of company manpower.

Re:ATI

[Performer Guy]

I disagree, NVIDIA are very prompt with their driver support, and they include the entire hardware interface, so they don't have to merge with something like the DRI.

Re:ATI

[cobar]

The guys at the Nvidia petition keep track of a few guys trying to port the drivers to FreeBSD. Unfortunately there haven't been many updates lately and I got tired of waiting and installed Linux till the drivers are ready. Too much fiddling in order to get a Voodoo 5 to work.

Re:ATI

[Performer Guy]

Most of those professional cards had no commitment from the IHVs many of the 'pro' drivers are half assed and created by throwing money and some specs at a 3rd party driver developer, they are ALWAYS in a distant second place to their Windows efforts. NVIDIA have great drivers for a four main reasons I'd say. Firstly they got of to a good start with SGI helping them before SGI bailed on yet another plan, secondly the are cash rich and able to spend the resources, thirdly and probably most importantly they reuse a MOST of the driver code from their windows code base and finally they seem to have a genuine appreciation that Linux support is a good strategic move for their business (for whatever reason).

Re:ATI

[Performer Guy]

Laptop driver support is uniquely problematic because the hardware & firmware varies significantly from OEM to OEM. Getting recent driver support on Laptops is problematic even on Windows.

Re:ATI

[darkonc]

it is colsed sourc so unless NVIDIA is willing enough

This points to the heart of the difference between open-source and closed source.

Closed source anything is at the mercy of the source vendor's business model. If and when what you want to do is consistent with what the vendor makes money allowing you to do, then closed source is fine. If/when what you want to do is inconsistent with the vendor's business model (unless they don't realize it), then you're pretty much SOL.

Open source allows development to be at the whim of the user base. It's pretty common that buyer interests and seller interests are not the same. Those differences also tend to increase as the stuff being sold gets older and older.

Re:ATI

[Skapare]

While I don't have any experiences using NVIDIA drivers, I have heard several horror stories about the inability to make certain system changes, such as upgrading a kernel to a new version due to some bug that makes this necessary, and finding that the driver being used croaks the whole machine. I don't remember if any of those stories involved NVIDIA drivers specifically, but they did involve video drivers with a kernel component that was NOT open source (e.g. was a binary-only module). Perhaps it could be the case that NVIDIA makes flawless code, and compiles a version of their driver for every architecture using the bus they support, and for every kernel and X version within 24 hours after they are released (something that can be easily accomplished by tracking the -rc versions every day). Perhaps they do this. Perhaps not. But there is value in having the source code so that you can choose to fix the driver yourself. Often times it is a simple fix, like merely recompiling.

I will always prefer the open source.

My form of dissing NVIDIA is simply to tell the truth about how they make their driver available, and explaining the advantages of having a community of many people doing the development, inspection, and testing ... and by NOT buying anything made by NVIDIA.

Re:ATI

[Performer Guy]

I think you are in the extreme minority.

Re:ATI

[Performer Guy]

P.S.

The original question (and article) was about hardware manufacturers who support Linux, nothing in there about BSD.

Re:ATI

[Paul Komarek]

I believe that HP worked with Pixar on one of the professional cards, maybe the FireGL but I'm fairly sure not any NVidia product, and I believe the resulting driver source was made available. I expect that Pixar went into this with their eyes open, and wouldn't accept half-assed drivers. But then, maybe Pixar is a half-baked company run by monkies that has just gotten lucky. ;-)

I don't mean to start a flame-war between ATI and NVidia (clearly I'm too late to *start* one anyway ;-). I simply expect that the professionals using accellerated OpenGL + Linux for animation or flight simulation (or other apps I haven't heard of) are not using consumer level cards with internal accuracy less than or equal to a C float. From that supposition, I concluded that there are 3D cards with accelleration under Linux which have high-quality OpenGL drivers.

-Paul Komarek

Re:ATI

[Fluffy the Cat]

Think there are ONLY binary drivers?

Yes, if you want 3D support.

You can download source for the NVIDIA GLX

I assume you're talking about the source RPM they provide? Sure, it's a source RPM (in that you can't install it directly and it lets you build a binary RPM). It doesn't actually have any source, though. Everything in it is a binary, and the Makefile's only target is "install".

the actual kernel module

You're right, we do have some source here. It gets built and then linked against a file called Module-nvkernel to give the final kernel module. Module-nvkernel doesn't have any source included with it. It's almost a megabyte of something that file claims is

Module-nvkernel: ELF 32-bit LSB relocatable, Intel 80386, version 1 (SYSV), not stripped

See that "Intel 80386" bit there? Oddly enough, that's not going to run terribly well on an Alpha. We have 280K of source which is possibly portable to other architectures and a 1MB binary file that's required and which isn't in any way portable to other architectures.

Its as simple as make and wowee, you have a brand spanking new driver compiled against your kernel headers.

Yes, as long as you're running on x86. Which Alpha isn't. If you're going to try to correct me, could you please put some effort into checking your facts beforehand?

Re:ATI

[Dwonis]

nVidia's drivers are crap. They're always behind, and the last I used them (before I dumped my NV card for a real one due to too many hassles), they didn't even support DRI.

Re:ATI

[Dwonis]

That's a nice excuse now, but when software -- especially operating systems -- becomes a commodity (and it's already happening), that kind of attitude will lose you your job or your company.

Re:ATI

[Dwonis]

The delay is still there. You just get a white screen, rather than a white screen with "NVIDIA" on it.

Re:ATI

[Dwonis]

ATI only releases a portion of the specs to a small group of developers who have signed an NDA.

It's a click-through NDA that anyone can get (unless there's some OTHER NDA that you're talking about). I got the access to the specs in about a week after I registered.

Re:ATI

[Paul Komarek]

Oops, my bad. s/pixar/dreamworks/.

-Paul Komarek

Note:

[Telastyn]

That just because they were community developed, doesn't mean the company didn't give out specs and info to facilatate the community's work.

3com cards seem to work on everything
Recent Intel network gear
Recent Nvidia
3dfx used to
IBM (even before the Linux money, their laptops worked well)

Re:Note:

[Afrosheen]

3com 3c905's and 3c509's have a tendency to choke/act weird/not work at all under Mandrake 8/8.1/8.2. Better off with a $10 dlink or netgear card with the realtek chipset.

Matrox and Nvidia

[mrfuzzee]

Matrox has the new drivers out for Xfree86 which work well, and a hell of a lot better then AcceleratedX. Nvidia also has drivers for Xfree86, and just kicks butt. I have been happy with both, They are relatively easy to install and configure.

Creative Labs

[Kaypro]

I believe Creative has a dedicated site for the continued development of their sound card drivers. They even have a CVS up as well.

http://opensource.creative.com/

Cheers!

Re:Creative Labs

[Mike Hicks]

Yeah, but they don't give out all of their specs. From what I understand, some info is given out, but a lot of stuff has to be reverse-engineered..

Zero marks for

[BrokenHalo]

UMAX - probably the worst supported scanners under Linux - I've got an Astra 610P, and still have to use WINE to get it to work :-

Re:Zero marks for

[Amazing Quantum Man]

Their USB scanners aren't supported either. I've got a 2100U... Come to think of it, the Win2K drivers for the thing are kludgy too...

Re:Zero marks for

[AJWM]

That's what you get for trying to run an input device on a printer port.

My UMAX Astra 600S -- the SCSI port version -- works just fine under Linux, never had a problem with it. (I don't use the cheap scanner-only SCSI card that came with it though, it's on a cheap AHA1540-clone card instead.)

Linksys

[misfit13b]

They've been pretty Linux-friendly in my experience for my home networking...

Re:Linksys

[The Asmodeus]

What's more is the 4 port Cable/DSL router I bought came with a Turbo Linux install CD. I mean, it's one thing to ship drivers but including a CD of Linux is top's in my book!

Nvidia...

[ishark]

Bad as it may sound, since they don't provide the source to their drivers, they seem to work seriously in improving them. I've been using them since my old TNT2 card, and the big problems present at the beginning have faded away to give place to a full featured, fast and reliable thing. I've also had answers to my mails reporting problems, which is always nice.

Speed is now at the same level of Windows, features seem to be there as well (I don't remember if everything works at every resolution yet or no), and over time they have become stable enough to be used as primary XFree drivers (in the beginning I used them only when I needed openGL support).

Given their work on the driver, I'm willing to live with their closed-sourceness. It's when it doesn't work and I cannot look in it to fix that I become less tolerant....

Re:Nvidia...

[ZaMoose]

And the fact that all I had to do was add an

Option "TwinView"
Option "SecondMonitorHorizSync" "30 - 110"
Option "SecondMonitorVertRefresh" "50 - 160"
Option "TwinViewOrientation" "LeftOf"
Option "MetaModes" "1600x1200,1600x1200; 1600x1200,NULL"

To my XFree86Config-4 to enable duall-head configuration pleases me to no end.

X running at 3200x1200 on 19" and 22" monitors is just too sweet.

Now if only I could get the GNOME menu bars to extend across both desktops...

3ware

[wiwo]

3ware actively supports Linux as there a linux drivers on the CD you get with their RAID-Cards. Works fine, at least with SuSE 7.2+

nividia and PCtel

[GutBomb]

I would have to say nvidia. they don't provide open source drivers, but usually their windows & linux driver updates are released at the same time, and actually right now, thier linux drivers are a bit more current then the official windows ones. (i am running the 28.80's in linux, but nvidia has only released 28.30 i think for windows) If i have to name another besides windows, I would have to say PCTEL. back in the days when NO winmodems worked, they had linux kernel modules for thier modems, even obscure onboard ones. I haven't heard much from them lately however.

Quite a bit, actually.

[awptic]

I've been subscribed to the linux kernel mailing list for some time, and there's quite a bit of discussion
coming from employee's of many popular hardware companies. NEC, Promise, IBM, SGI, SUN, to name a few.
Then there's the ever so popular drivers developed by NVIDIA, closed source unfortunately, but that's
a company policy iirc.

Agere and 3Com

[scumdamn]

Both Agere and 3Com have drivers available that they've written.
I don't know the quality of either, but from what I hear, Agere's drivers are good for linux. I know they are for other operating systems.

Matrox

[shaldannon]

Matrox is actively supporting its line of dual-head cards under Linux and various flavors of Windows. There may also be *bsd support as well, but not being a bsd user, I didn't pay attention. I'm running a Matrox G450 under Red Hat 7.2 (upgraded from 7.1) with two ViewSonic E771 17" by .26 monitors in merged display mode and it is phenomenal. I had to use their tech support list to get it working, but it only took a few days...mostly because I'd ask the question from work, try the solution at home, and then follow up at work. See the screen shot (2560x1024 .jpg image, 10485992 bytes).

Compaq

[BayStealth]

I am in the process of bringing our brand new network on-line (8 new DL-360s) and Compaq has been extreamly helpful. All of the servers are running RH 7.2 (they were delivered with 7.1 installed) and we have run into several issues reguarding RAID, the LightsOut boards, etc. Compaq support for their hardware and software has been excellent. Not to mention several cool software things that came with the servers.

Re:Compaq

[A Masquerade]

Compaq kit works OK if you use vendor (ie stock RH) kernels. The RAID controllers are well supported in mainline kernels.

However the on-board monitoring requires binary closed source modules which crap all over your kernel if its anything other than RH stock - so if (for example) you want some extra drivers and FreeSWAN then using Compaqs monitoring modules you can crash the machine in 3 seconds purely by cat-ing a few /proc files!

The installer is interesting as well - it "fixes up" the modules to load with the current kernel - forget ABI compatibility - this forces modules into an incompatible kernel and modifies the version stamps so you don't notice.

And compaq's packaging stuff is shit - why does no one have any idea how to make rpms - these do really stupid things. The only worse packager than compaq are ibm who have binary rpms which build new rpms on install!

Basically never trust any vendor that uses ServerWorks chipsets (ie both Compaq and IBM) - the big NDAs surrounding those chipsets will come back and bite you hard!

For USB scanners: Epson

[Basje]

According to the sane USB page they release even preliminary specs on demand: http://www.buzzard.org.uk/jonathan/scanners-usb.ht ml.

Mandrake linux detected my 640U flawlessly, and it works great. And on top of that, it scans better and faster than my old scanner, which I killed while trying to get it working under linux :) (which I shall not name here)

Re:For USB scanners: Epson

[larien]

Yup, got an Epson scanner which took me a little while to get working, but nothing difficult; just that I'd never used Sane before. I now do my scanning in linux rather than Windows, as I can understand Xsane better than the Windows software which came with the scanner.

However, watch out; one scanner (the 1250) doesn't work under linux. Check out the link from the previous article for a complete rundown of supported printers and how well they work before you buy!

Many do..

[psavo]

• Well..
• Matrox
• nvidia
• intel
• ibar (a.k.a ibm ;)
• HP (deskjet printers)
• OKI (4w driver was sponsored by them)
• AMD
• ATI (sortof. at least their linux drivers sucks as much as windows one..)
• ... pretty much more.

Jeesus christ this lameness filter gets my ass. no wonder there's THGSB week going on. This is SO lame.

Re:Many do..

[lkaos]

I've got an ATI Radeon in my machine, my gf has an ATI Radeon in her Windows 2000 box, and I can confidently say that the Linux drivers are far better than those for Windows, probably because ATI didn't write them. ATI (unlike nVidia) have been very good about releasing specs to the community.

This is because the special ATI hardware optimizations integrate better with X than they do with Windows. Of course, this is only possible with > X4 since that's when all the XVideo stuff was introduced.

BTW: ATI does not provide specs to the community. They provide only a portion of their specifications to a very small number of developers who have signed an NDA. The aspects of their cards that they feel are important to their bussiness model (i.e. TV-Out, hardware iDCT) they have released nothing for.

Re:Many do..

[psavo]

I decided that it would be better to reply to your mis-information than simply mod you down.. I'll leave that job to somebody else.. :)
Go ahead. that's what M2 is for.
The question was about active support for Linux. I consider writing the driver as one.

nvidia

Bzzt.. wrong! nVidia provide binary-only drivers. There are stability issues, and there's no way they're going to be resolved, because no-one's got the source to fix it.
So how do you think Windows issues are resolved? By peerin' over source, huh?

BTW, I own a GF2MXPCI and nvidias drivers DID crash my computer evry 3hours. that was about 3 releases ago, I use now nv driver instead, I'm not a gamer.

And remember, there's a hellot more to nvidia than simple graphics. They're into chipsets now, too. It ships OSS as of now.

Re:Many do..

[ninewands]

nvidia

Bzzt.. wrong! nVidia provide binary-only drivers. There are stability issues, and there's no way they're going to be resolved, because no-one's got the source to fix it.

Bzzzzt ... wrong!

Sources for both the kernel driver and the GLX driver are available from Nvidia's linux download page in both SRPM and tarball format. This is necessary because NVidia cannot possibly provide binaries for all possible combinations of distro/kernel (the nvidia drivers have to be compiled for a specific kernel version) that might exist, especially since there exists such things as LFS, Rock Linux and Sorcerer Linux.

As for fixing the stability problems ... well, I guess you are free to do so, but the drivers are NOT "free software". Technically, I guess they are not REALLY Open Source either since you can't re-distribute the modified driver, but I imagine nvidia might look with favor on receiving a contributed patch, provided it didn't break something else.

Re:Many do..

[dinivin]

You really need to get a clue.

The SRPM and tarballs that contiain the GLX driver "source" don't contain the GLX driver source. And the kernel source simply provides hooks for the driver into the kernel, and contains none of the functionality for the card.

Dinivin

Re:Many do..

[Velox_SwiftFox]

There's one company everyone seems to be missing out here: Adaptec. They took over maintenance of the SCSI drivers for their cards in the Linux kernel a while ago, and they've put a lot of work into them.

Unfortunately, I have to deal with some systems that have Adaptec RAID cards installed. They haven't released anything sourcewise since their "unsupported" 25Jul2001 patches for the I2O-based controllers.

Someone should inform them that there is a general-purpose kernel distribution past version 2.4.6 too...

3ware...

[MonkeyBot]

I have several 3ware raid cards that have worked great. Not only that, but I've had to call several times for support, and every time, I either talked to someone who helped me right off the bat, or was contacted by someone who knew what they were talking about within the day. Twice, they even made driver fixes on the fly and sent me the updated code the next day. DEFINITELY the best company-based Linux support I have seen...

hit and miss...

[laserjet]

Linux support is kind of hit and miss right now with larger companies. For instance, HP is adding more linux support than ever to their printers, even the office jets, but if you go buy a scanner, they don't support it. Obviously, the community supports a lot of HP scanners, but not the company.

HP is also supporting RedHat on it's new Itanium servers, and also supports RedHat with its mid-range storage arrays. They seem to be testing the waters, and I think they are doing all right for such a large and slow moving company.

Samsung is also supporting their printers, by offering Linux drivers and Linux phone support (minimal, but it is there). This is a good thing.

Qlogic and Emulex both support linux with some of their fibre channel HBA's.


So as you can see, you kind of have to pick and choose who you get our stuff from. The corporations are still in the "test the waters" phase for the most part, before they dive in to linux head first. They don't want to get burned by wasting money doing all the work if it will not pay off. In another 3 years, I think Linux support will be fairly mainstream as far as business server and workstation equipment go, but it may still be hit and miss in the consumer market (i.e. webcams, cheap USB scanners, cheapo printers, etc.)

Re:hit and miss...

[Lumpy]

the problem is not the company as a whole but the subcompanies it is made of.

the scanner division is not really HP. it's still the origional company and ran the same way it was when it was bought by HP. the scanner people still act like they are holier than god and that linux is evil and they wont waste any time with such as an obscure operating system. (Their attitude not mine) and they think the off the shelf chipset they used in today's usb scanner is a company secret and HP in it's entierity will collapse if you discover what it is..

the printer guys started life when you onyly had Unix and they look at linux as a welcome draw back to the Unix roots... same with trhe server guys...

it's all the microcosims insode the corperate whole.. it will take either mass firings or a iron fist from above to get the scanner group on the linux bandwagon... and the CEO/CTO/CFO/EIEIO dont give a rats ass about anything other than seeing the Profits line move upwards.

DLink and a noname laptop

[NETHED]

DLink has pretty good support, especially for Linux. My dad's noname laptop came with a CD that provided Linux drivers, and they actively support them via the phone support.

Gigabyte has been good to me.

[SwellJoe]

I don't know if it is universally the practice at Gigabyte, but the networking server group there have been great. They've always made sure up-to-date drivers shipped with everything I've gotten from them. Some of their boards ship with Promise ataraid controllers, and while they couldn't get me the docs they tell me the techs there have been campaigning Promise to be more forthcoming (and they do provide binary drivers for those controllers--I don't use them, but they are there and Gigabyte actually apologized for not having source drivers available).

Matrox seems to be good too, as I've never had trouble getting their video boards to work right out of the box with X (as I understand it the Matrox folks are more helpful than most to the X developers).

That said, Promise is clearly bad for refusing to release their drivers in source form (I guess they think their software RAID technology is so advanced it would give their competitors a great benefit--or maybe they are embarassed to let us see it). Logitech have never been friendly to the OSS world about their QuickCam cameras. I think a lot of printer manufacturers have been a nuisance in this regard (I gave up on trying to figure it all out and bought a Postscript-capable network printer). I'd be curious about good and well supported inkjet printers, though...

Oh, yeah, our Microtek X6EL scanner works great with Linux and SANE. I don't know if the manufacturer is to be credited partially or if the driver author was just heroic in his efforts, but it works exceedingly well.

Re:Gigabyte has been good to me.

[ncc74656]

As far as inkjet printers go, you can't go wrong with Epson.

...as long as you don't mind the stupid POS clogging its nozzles after a few months. :-P (This was a constant problem with the demo printers back when I worked for The Man.)

Others have noted that Lexmark provides halfway-decent Linux support for its printers. I have an Optra Color 40; while it doesn't need Ghostscript as a print processor (it groks PostScript all by itself), Lexmark provides source code for a command-line utility to swap cartridges and do other maintenance tasks. (Others have made WIMP tools that do the same.) As for the other models (the ones that don't even do PCL, let alone PostScript), others have said various models work as well. I don't have any experience with them, though. It's also worth noting that if the nozzles get clogged beyond hope of unclogging, you don't have to send the printer in for repair...you buy a new cartridge for $25, pop it in, and get back to work.

There are quite a few !

[forged]

Pick one source from the following list, in no particular order:

RedHat Hardware Channels
http://www.redhat.com/marketplace/channel_hardware . tml
(among others, there are Dell, Egenera ..)

Linux Hardware
http://www.linuxhardware.org/

Linux at IBM
http://www-1.ibm.com/linux/

Linux at Compaq
http://www.compaq.com/products/software/linux/

It is a safe assumption that hardware from the 2 above manufacturer will be well supported, since they are supporting Linux heavilly.

Last but not least, make sure to read the Howto:
Linux Hardware compatibility HOWTO http://www.tldp.org/HOWTO/Hardware-HOWTO/

Re:Typical response

[swagr]

This happened to me during an ADSL support call. They switched their DNSs and I wanted to know the new ones.

"..I'm running on Linux..."

"O.K. Go to Start... Settings... Control Panel..."

"No. I'm not running Windows, I use Linux".

"On a Mac?"

"I just need to know the DNS numbers."

"O.K. What's the problem again?"

"My connection has been working fine. I ping IP addresses but can't resolve domain names. I think you guys switched your DNSs IPs."

"......"

"Do you have some numbers beside something that says 'DNS' or 'Domain Name Server'"

"....... Oh yes."

"Can I have them."

...

BusLogic (Mylex)

[MoNsTeR]

Do they still exist?
Anyway, I remember they wrote all their own linux drivers for their scsi cards...

Lynksys

[Technician]

The 3 port standalone print server has excellent drivers and docs. They didn't write the drivers, and do not claim to support them, but they did an excellent job of finding them and including them on the CD. The software is provided as-is as they state they don't support it. They do provide docs on server interface and how to connect and configure it. You can even FTP a print job to an attached printer.
If you want to share a dot matrix, laser, and inkjet with your Linux/Win mix LAN, this is a good way to go. TCP and several other network protocols are supported and can be enabled/disabled per your needs. It does not provide spooling. A machine configured to spool the jobs will be needed if you desire this feature. Otherwise the printers appear (and function) as local printers via the driver. 2 of the 3 ports support bi-directional centronics printers.

NVIDIA For One....

[CDWert]

I would say NVIDIA for one, people complain constantly about closed specs etc.

But the truth is it would be competivley BAD for Nvidia to release the specs, yes others have, they choose not to, thats fine with me, they do provide GOOD drivers, and the SRPMS, as well as tared gzipped kernel modules for you to compile on any Linux setup you wish, the actually libs are closed source but hell they DO provide drivers for an OS that accounts for a VERY small portion of their sales market.

There are other vendors that provide Linux support, to be honest If I was in charge of a HW company, I wouldnt, I would provide the specs under some kind of closed agreement to 3rd party developers.

NVIDIA Does provide nice linux drivers, I have, unlike other never had any problem, they release newer version and each generation (for the most part) they get better what more can you ask....(and please dont say provide the specs, if you are thinking or saying that Im betting you have no experince in engineering hardware for a commercial market where competition, especially in th 3d accel, is just downright evil)

Re:NVIDIA For One....

[Dominic_Mazzoni]

I don't quite follow your argument. Yes, it's good that they've provided binary drivers - better than nothing, right? And they do seem to keep them up to date.

But I have two NVIDIA cards to support dual heads. Unfortunately their driver crashes immediately if you try to use it on both cards simultaneously. The only way I can use two heads is to run NVIDIA's driver on one card (which gives me video acceleration) and the old, open-source driver on the other card, which gives me no acceleration.

So while I appreciate NVIDIA is trying, their drivers are not perfect, and thus they should either open up the source, or the specs.

Re:NVIDIA For One....

[Kynde]

NVIDIA Does provide nice linux drivers, I have, unlike other never had any problem, they release newer version and each generation (for the most part) they get better what more can you ask....

Sources for the kernel module, period. They could keep their userspace part of the driver as closed as they'd want, but the part that gets loaded to the kernel space should, without questions nor excuses, be open source. Doesnt have to GPL, but open, so that kernel developers can see what it does and take action into stabilizing it.

There just simply aren't any excuses for keeping their driver binary-only. The only ones that I can come up with is either, terribly hacked-kludged-bastardized hebrew code that would only give them bad name OR fucked-up management that fails to see the benefits in opening up the code.

I'd bet for the later, being the common case in almost every company is most likely the cause of our distress.

The current situation is that kernel developers cannot do anything about it. When a kernel that's running with the NVidia module acts weird, the only party that can do anything is NVidia and quite frankly it isn't enough.

Re:NVIDIA For One....

[CDWert]

"There just simply aren't any excuses for keeping their driver binary-only. The only ones that I can come up with is either, terribly hacked-kludged-bastardized hebrew code that would only give them bad name OR fucked-up management that fails to see the benefits in opening up the code."

It isnt. I hasnt been for damm near a Year now, the source is there go download it.

The source to the kernel module is open.

The userspace part is in binary only.

I think most people, and I have been wondering this for some time , since I looked through their sources. Have looked latley, or actually even know what they are looking for...

I was going to post links to the sources, but the damm Nvidia site is down (Yes they are there)from here.....If you need em email me....

Re:NVIDIA For One....

[CDWert]

True to some extent.

But NVIDIA IS Doing pretty amazing things at a driver level, they have a unified driver architecture, speaking in common with ALL their hardware, a blah chip from last year and a supernew Ti??? speak the same to the driver. They have some other name for what I speak of, but from a driver level, having my TNT2 driver be the SAME as my Ti400 driver says something pretty spectacular about the very architecture itself. I bet ATI would love to have all the specs on it, and IM betting 10 to 1 if Nvidia released the source the would.

Good to hear

[tweakt]

I just bought a bunch of new gear and built a server with the intent of setting up a 3-IDE drive RAID5 software under linux-2.4. Well as luck would have it, some wierd bug has bitten my system and I'm getting the dreaded PCI timeouts which hang the whole thing solid. I've tried a bunch of stuff and decided it isn't worth my time to try and solve.

So I've given up on that and ordered a 3ware 6410 for $99. True hardware IDE RAID5 for under $100...not bad. Good to hear they excel at support. We'll see how it goes when I get it in a few days. *eagerly awaits*. I especially like the fact you can download a full source driver tarball from their website. But of course the driver has been in the kernel since mid-2.2 days.

Snap up those 6000 series, it looks like they are discontinued (and 7000's start at $250 and are 64bit only! ack!).

Manufacturers may help, but...

[fruey]

In general, the support for hardware can be guesstimated:-

1. If the hardware is top end, and likely to be owned by Linux people (gaming graphics cards, hotrod modems, cool peripherals) then they are fairly likely to work, with obvious super-high end exceptions. Top end hardware also usually follows specs for standard stuff (like standard SVGA, etc)
2. If hardware is low end, forget it. Most of it is manufactured in bulk for Windows only, may have some proprietary code where standards would have done, and is less likely to be owned by a Linuxer anyway. Exceptions below*
3. Latest products : unlikely to work because drivers won't have had time to get integrated into kernel development, however modules may be available. Again, if it follows standards then it may work (with performance hit) with generic drivers anyway

I have seen 3Com mentioned, well there's a case in point where they are industry standard network card people. Loads of people have 3Com cards. Loads of people having certain hardware means it's likely to be supported, however....

* Very popular shitty low end hardware may work due to good hacks by lots of owners, however reverse engineering isn't an exact science and strange hardware stuff means only hardware which is technically acceptable in it's I/O style will work.

Manufacturers who only develop for Windows are most likely to be found having market share in low end products. The top class lot are much more likely to work. Peripherals that are little more than I/O ports which are instruction driven from host processing (huge binary drivers required) won't work with Linux unless the manufacturer releases all the specs.

I would say that manufacturers make regular business judgements on all their support: because Linux doesn't have market share enough to make it a sales point to support "end user" hardware and they won't release code (because competitors making low end shit will steal it and obfuscate it as a Windows locked binary) but server hardware is supported rather more quickly, because the server market share for Linux is substantial enough.

Re:Manufacturers may help, but...

[AJWM]

You seem to have a different definition of "top end".

I've seen a ton of top end stuff for which only NT (or W2K) support exists or will exist. Examples: high-end (broadcast quality) video capture/output cards from the likes of DPS or Matrox, hardware RAID cards, high-end audio cards, etc, etc. Perhaps this is "super high end" to you since these cards are typically a couple thousand dollars each. Certainly I wouldn't by them new for a Linux box (no support), but when I have a roomful of boxes with DPS Perception composite video cards that I can't migrate to Linux, it's a bit frustrating.

Watch for drivers being GPL, but not in kernel

[Yohahn]

With the advent of hotplug, and firmware uploading, there are going to be lots of firms offering support for linux, but it won't be included with the kernel.

I work for a company that will be releasing firmware for our devices, and a script that makes it work with hotplug. We can GPL.

I worry that drivers like these won't get the attention that ones in the kernel do because they aren't included.

I hope that there will be some common method of installing firmwares or a commmon repository of firmwares in the future.

Linux users seem to depend on drivers being included with the kernel, having nothing else to get.

Re:Watch for drivers being GPL, but not in kernel

[Wavicle]

That's a fundamental problem with monolithic kernels. Linux will continue to be rejected by the unwashed masses when adding new hardware means compiling a new kernel.

The "release often, release early" mantra is no longer used with kernel development. If your company released "Cool Hardware Thing" on March 1 and sent open source drivers for it to the community the same day, Joe Average would still not be able to use your hardware because 2.4.18 was released before then and 2.4.19 still isn't out (I don't see Joe Average being comfortable patching and recompiling his kernel). The situation is actually worse than that since Joe Average would probably have to wait for Red Hat to offer up that kernel revision it could easily take forever (or maybe the 2.4.18 kernel is available for users of 5.0 ?)

I have to admit, it's pretty slick the way windows comes up and says "Hey, I see you added some new hardware, why don't you stick in the CD the manufacturer gave you and I'll make it work".

</favorite rant>

Oops

[tweakt]

Looks like "hit and miss" is missing an ;-)

Wireless cards

[xiaix]

When I went looking for wireless cards for my Vaio running linux, I found as the author did that most of the 'support' for linux means 'some one figured it out'. Although this is part of the beauty of being a linux user, sometimes you want to know that the hardware manufacturer actually knows you are out there and cares enogh to support you. In the end I wound up buying a Cisco Aironet 350 card ($125 bux at computers4sure.com), which came with linux drivers, software, and install instructions.

I dont mind spending a few dollars more to support a company/product that supports my choice to use linux. It was well worth the extra $ to plug it in, run the install, and connect to the network at my college in under 5 minutes.

AdvanSys

[Tet]

Now part of ConnectCom, and marketed under the "AdvanSys by Initio" brand. Not only is the advansys driver in the kernel written and actively maintained by the company, but it's superb quality, as well. By far the best SCSI controller I've used under Linux, and I can't recommend them highly enough.

Adaptec, Belkin and !Creative

[entrigant]

I'm surprised no one has mentioned it but Adaptec has done a very fine job supporting linux. I am not sure how many if any of the drivers they actually wrote but they have a really well designed web site to help linux and *BSD users setup and use a lot of their equipment. They also provide utility software for their hardware. For example I am running Adaptec Storage Manager right now on a linux system with an Adaptec 2400a raid card.

Belkin also does many of the same things. I know that belkin has a rather wide variety of hardware they sell, however with their UPS's I know for sure that linux is very well supported. Their upsd and ups monitor are closed source but they work very well. They are also rather well documented.

There is one company that really bugs me though and that is creative. They have opensource.creative.com. They've made many announcments and claim bragging rights for supporting the linux community. The truth is however every driver for a creative device out there has been written by the community with barely any input from creative. On the emu10k1-audigy driver mailing list there's a guy.. I forget his name.. who works for creative that does get info from time to time for the development team, but it always seems like he has to beg or plea for the info he wants to get. Usuaully it seems as if he just asks someone who is coding the windows driver or helped design the hardware without getting approval first from management. I'm not implying anything here other than creative is not actively supporting crap.

The community writing drivers is their support

[SuperguyA1]

If we want to live in a non-mono OS society it could potenitally be impossible or at least unrealistic to think that every hardware manufacturer would write drivers to every OS. The fact that the Linux community was able to write drivers means the company probably opened up their hardware specs. This in its self is a HUGE help, not only allowing LINUX drivers but *BSD, BeOS(*sigh*), Plan 9 or whatever to have drivers as well.

Re:HP

[iCharles]

With respect to the Enterprise market, HP is really making a push in the Linux market. They regard it as one of the three primary OSes for their high-end Itanium systems (Win64 and HP-UX are the other two), and are integrating it into their managemnet and partitioning tools. Further, APIs and other software has been created to improve interoperitability between HP-UX and Linux, so that software for one will work on both with minimal effort. Finally, the new "Blade" currently only support Linux (though they are planning to support Windows in the future).

Adaptec

[Paul the Bold]

Adaptec is very cooperative. You can find their page here. I think the aic7xxx driver in the 2.4 kernel tree was sponsored by Adaptec (i.e. they paid a guy to write it). It works very well. Here is the official page for the aic7xxx driver.

Re:Adaptec

[shlong]

Adaptec pays people to support the following products on Linux (and FreeBSD, to a lesser extent):

• SCSI HBA
  
• ASR and AAC RAID
  
• External RAID
  
• Fibre Channel (not open source, due to licensing with Agilent)
  

Re:Adaptec

[frleong]

Of course Adaptec is supportive of Linux. Linux market share in server systems is rather high and well, many servers need SCSI to operate efficiently. In fact, I am pretty confident that most (if not all) server hardware vendors support Linux (network cards, network management tools, SCSI adapters, UPS...)

Re:Adaptec

[cobar]

For what it's worth I believe they hired one FreeBSD developer to continue supporting the drivers he wrote. Can't remember his name off the top of my head.

Re:Adaptec

[shlong]

They hired three, but we are officially paid to work on Linux. We just 'prototype' our work on FreeBSD =-)

Btw, the name that you are thinking of is Justin Gibbs.

Lexmark puts a Penguin on their Boxes

[Holesome]

I just bought a cheap, er, inexpensive Lexmark laser printer. It touted Linux support and even had a Penguin on the box and linux drivers on the installation CD. Unfortunately it still took a bit of fiddling to get the printer to work but work it does. Can't say to what extent they actively contributed versus used other peoples work however.

Re:HP

[CJ Hooknose]

I bought their cheapest (stupid move) laser printer a few months back and it doesn't like linux (Only supports PCL not PostScript).

? Printers don't have to speak PostScript. It makes things easier when they do, since everything speaks PostScript under Unix, but there are PostScript->(printer native data format) converters out there. They typically go by the name "print filters", and every distro includes a metric arseload of them.

PCL is pretty standard, so PostScript->PCL print filters are mature and stable. Your distro's setup tool (YaST, "setup", DrakConf, linuxconf, whatever) almost certainly has a "setup printer" option in it. Use this if you can--it's generally pretty easy, and involves letting the parport auto-detection work, or picking your printer from a list of models. If that doesn't work, try going to linuxprinting.org , entering your printer's model# into the search box, and following the directions.

UMAX

[ninewands]

The problem with UMAX CSCI scanners is the crapware semi-SCSI interface card they provide with them. Replace the card with an Adaptec, or some other supported REAL SCSI host adapter and you will find that the UMAX scanners are very nicely supported by SANE.

False statements

[jmv]

I wonder how many people have encountered the problem I have right now. I bought a Midiman Delta 1010LT, which the company claimed worked under Linux (through "third drivers", from ALSA). This was fine with me so I bought the card (which box had a nice linux sticker on it).

I then tried to make the card work under Linux, only to find out that it wasn't supported by ALSA and that though there were some efforts under way, AFAICT nobody has ever been able to output a single sound out of that card. I wonder how many companies use this kind of false publicity with Liunx.

Re:False statements

[jmv]

Last time I checked, I think the Audiophile 2496 was supported by ALSA (probably a couple months after you bought it). I'm guessig that's the way they operate... They release a card and put a Linux sticker on the box that means (to them): "it would be nice is someone wrote a Linux driver for that card" and then they wait until someone finally writes a driver... Not to mention that they don't even respond to e-mail asking for the card spec sheets.

Re:False statements

[jmv]

Try the latest ALSA beta (0.9.2 I think) or the ALSA CVS. I've seen the source of the driver and there seems to be all that's needed for your card. Then again, I don't know how tested the driver is.

Re:Typical response

[digitalsushi]

I asked for these from a girl at college a couple years ago, the day we were moving in, and she went balistic on me and told me to get my own. She was CS, too.. sad..

Project UDI

[Deven]

I suppose it's reasonable to expect a company to produce drivers for Linux, but remember, there are umpteen billion operating systems out there, and these companies don't have the time or resources to develop for all of them.

That's why we should all be supporting Project UDI (Uniform Driver Interface). You write a hardware driver once and it works (unchanged) on all UDI-enabled operating systems. What could be better?

Re:Project UDI

[4of12]

Sounds like a really good idea.

But there are two sides to an interface: the driver and the OS.

How well can UDI cope with the ferment that is the Linux kernel: things changing to make them faster, cleaner, etc.

Re:Project UDI

[twoflower]

That's why we should all be supporting Project UDI [project-udi.org] (Uniform Driver Interface). You write a hardware driver once and it works (unchanged) on all UDI-enabled operating systems.

At the price of absolutely terrible performance and scalability. UDI support has been soundly rejected by Linus Torvalds, Alan Cox, Peter Anvin, et al every time it's been brought up.

It's not just Linux developers who think the idea sucks, either -- that's why it's not supported by other OSes as well.

Twoflower

Re:Project UDI

[AJWM]

Well, it's the old portability vs performance tradeoff.

Sure, a native driver tuned to the OS is going to give you the best performance. But a portable (eg UDI) driver is going to give you better performance than no driver, which is what you'll have if the hardware vendor hasn't chosen to support your particular platform.

Sure, I'd rather have native Linux drivers for the various gadgets I own. But some of those gadgets only ever came with Windows drivers -- I'll take a UDI driver over the no-driver-at-all that I currently have for those.

Re:Project UDI

[CAIMLAS]

Urg... that's another level of abstraction I don't need. I'd take my performance on the hardware that isn't exactly my choice, over poor quality performance on the hardware I want.

I might as well use an OS written in Java... (no offense to java, but that'd suck.)

Plus, I'd think it would be more difficult to write a UDI driver that's stable and bugless than one for an OS - once again, there's a level of abstraction. Sure, though, if it could be done well (adn I doubt it could be) it would be nice to see such a thing done by manufactuers of devices (like portables, etc - not video cards and essential stuff like that. that = relaly shitty performance) so alternative OSes could have support for nice devices like that.

Re:Project UDI

[Deven]

But there are two sides to an interface: the driver and the OS.

How well can UDI cope with the ferment that is the Linux kernel: things changing to make them faster, cleaner, etc.

Everything OS-specific is encapsulated within the "UDI environment", which is the OS side of the UDI implementation. An environment needs to be created to allow an OS to use UDI drivers, and this might be relatively difficult to do at first, but it only needs to be done once for the OS.

When something changes in the OS requiring a change to the device drivers, that change occurs once in the UDI environment, rather than repeatedly for every relevant driver. That's the advantage of using a clean API...

Re:Project UDI

[Deven]

There are good, technical reasons dealing with maintenance going forward for not wanting to enable closed source drivers.

Could that be because so many Linux kernel changes have required mass updates of driver code to match? UDI is designed to abstract out all of that stuff into the UDI environment on the OS side. Old drivers should continue to work as well as they ever did, since the UDI interface is more stable than the Linux internals...

Re:Project UDI

[Deven]

At the price of absolutely terrible performance and scalability. UDI support has been soundly rejected by Linus Torvalds, Alan Cox, Peter Anvin, et al every time it's been brought up.

It doesn't seem like they've really given it a fair chance. From what I heard, they looked at it a little, decided it looked complicated (which it is) and presumed that complicated == slow (which is an unwarranted presumption), decided that the whole UDI thing was just a ploy by Intel to get Linux folks to do all their driver work for them (Intel was late to the game), and started railing against what a terrible idea UDI support was.

And, of course, once a determination like that has been made, everyone's going to be even more close-minded about evaluating it in the future, preferring to rely on their bad first impression than look closer at it.

I've heard the comment that a 5% loss in performance would be so terrible that it wouldn't be worth using UDI if it made things that "slow". Nevermind that native drivers could run alongside UDI drivers in the cases where performance really matters that much...

According to the UDI developers, they're getting equal or better performance on Unixware compared to native drivers, despite the extra overhead of layering the UDI environment on top of those same native drivers! It should be possible to make a more efficient implementation that is faster yet. The UDI design always had performance and scalability in mind.

It's not just Linux developers who think the idea sucks, either -- that's why it's not supported by other OSes as well.

Stop the presses! People don't want to chuck out legacy interfaces and have to learn something new? Who could have imagined that would happen?

Of course UDI has an adoption problem. Everyone wants to keep doing things the way that's always worked for them, regardless of what's best for themselves and the industry in the long run.

UDI offers 100% portability from one OS to the next. Plenty of Linux users have complained bitterly when no Linux driver is available for their hardware -- you think they wouldn't be happy with a working driver, even if it's a little slower than a native driver might be? Yes, we're starting to see some native Linux drivers written by hardware vendors, but more often they expect the Linux community to do the actual work anyhow.

What about other alternative operating systems? If every Linux driver was available as a UDI driver, those drivers could be used on FreeBSD, BeOS, GNU HURD, or any other operating system -- just write a UDI environment for that OS once and all the UDI drivers suddenly become available. Richard Stallman should jump at this, but instead he decided it was a ploy by Intel to leech free driver efforts from free software developers. Why should those developers waste time porting each driver to each OS? What a waste of talent!

Linux is in a stronger position than it used to be; hardware driver support isn't as comprehensive as for Windows, but it's probably #2 now. That doesn't really help HURD, FreeBSD, OpenBSD, NetBSD, BeOS, and every other alternative OS out there, and the closed-source Linux drivers out there couldn't even be ported. UDI drivers could run unmodified across a range of OS platforms.

UDI also offers many safety and security benefits that native Linux drivers don't have. Misbehaving UDI drivers could be killed and restarted by the OS without harming the system, if the environment implements such protections -- the drivers could even be implemented in user mode if desired. Hardware drivers are often blamed for system crashes, especially on Windows -- wouldn't it be great to be able to protect the system from any drivers which couldn't be trusted not to misbehave? (Performance-critical drivers could be configured not to be protected as much.)

There are a lot of good reasons to support UDI, and the main objections (performance, scalability, etc.) all seem to be based more on perception than actual experience with UDI. That's unfortunate, because the potential benefits are enormous, and everyone is too busy dismissing the idea out of hand to even try to make it work.

Don't we all know what it's like to be the underdogs in terms of hardware support? Many other alternative OS users are much worse off than Linux users in this. Must we forever have this morass of legacy driver interfaces and eschew fundamental advancements in the state of the art? (UDI is a clear advancement, at least in interoperability, and possibly in design as well.)

There's some pretty sharp people working on the Linux kernel. I bet they could make some highly efficient UDI implementations, if they wanted to. The design always had performance in mind (zero copy, DMA, etc.) but the prototypes and reference implementation were never highly performance-oriented. The portability of UDI has been proven, and even without a focus on performance, they've been able to rival the performance of native drivers. Imagine what the Linux kernel hackers could do if they did focus on performance in a UDI implementation?

Re:Project UDI

[alext]

I might as well use an OS written in Java... (no offense to java, but that'd suck.)

Well, probably, now... but actually it's not clear that this will be the case forever. Yes, I will point out that Java gets quite a bit faster with each release, but more importantly the hardware gets more diverse year on year. CPUs with big register files, vector operations, 64 bit operations... there's a fair slew of chips out there already, before Clawhammer & co appear, and I doubt if C compilers are going to optimize for all of them. In fact, that's impossible. So step forward the JIT, the guy that knows your hardware, and even your usage patterns and can optimize for both. It's the only practical solution longterm... convinced anyone?

Version incompatibility

[matze235]

I believe that one of the reasons why there aren't more Linux drivers is that drivers are incompatible with every new kernel release, so binary-only drivers don't make much sense. And binary-only drivers are the only way most vendors want to publish their drivers.

Thanks -

[wirefarm]

That's exactly the sort of answer I was looking for when I posed the question.
When I asked it, it was not "What works well under Linux" or even "Who makes drivers available" but "Who really stands out in their support of the Linux community."

At this point, I've pretty much got all the hardware I need, most of it working under Linux, now it's down to where should I go to upgrade and who do I recommend to friends?

I've had great luck with Adaptec - they make great stuff and I never had to give the drivers a thought - I just never knew that Adaptec was throwing so much support behind their product.

Cheers,
Jim in Tokyo

NDAs, DMCA, etc..

[Mike Hicks]

I think there are a lot of companies that would like to do better, but can't due to non-disclosure agreements of one kind or another. Video card vendors like Matrox and ATI can't give out all of the information on their cards due to Macrovision support on TV-output ports.

I have a laptop with an Intel chipset that has an integrated winmodem that I can't use. Intel is usually very very good about releasing specs (definitely something I'd say they're better at than AMD and Via), but due to proprietary technology, no specs are available, and I can't get the damn thing to work.

I always get confused when this happens. I always thought that the proprietary-ness of an object was contained within that object. Why companies are so scared to release info on how to get something to work is beyond me. I guess there are some decent reasons for the Macrovision problem (I hate the reasons (it's illegal in the US to not have Macrovision protection, AFAIK), but they are valid nonetheless).

I hope that Linux will pull some of these companies away from that line of thinking..

Anyway, I don't know if it's still true, but Epson used to release quite a bit of info about their printing languages. I think HP did as well, at least until they got into their winprinter phase. They seem to be loosening up.

Hmm.. I think that some of the best companies in this regard have low profiles. All of the big names I can think of have made some pretty poor choices, IMHO.. A lot of companies seem to want to release just enough information to keep Linux users happy.

I think it's best when companies release this information, though. When the specs are opened up, it means that the product can have a much longer life cycle. As long as there's someone who is interested in keeping a driver working, it'll work. I bet there's a bunch of stuff that's supported in Linux that doesn't work in Windows anymore..

Software modem manufacturer support

[Fluffy the Cat]

IBM have a driver for their MWave software modems in the main kernel tree. PCTel have an open driver for theirs, too. Conexant are actively supporting the development of drivers for their HCF and HSF chipsets. Lucent, on the other hand, have an unofficial binary driver that seems quite happy to crash.

My list...

[OneFix]

Well, for video...no contest, it's nVidia...

Which incidently, the RivaTV Project had a recent breakthrough which means most of the cards with TV-IN are now working.

Printers...HP...

Scanners...Epson...

What really sucks is I have an Optrox scanner...some will know, this company went out of business a while back, and there seems to be no specs available for it.

What I would really like to know is if anyone knows how one would go about getting specs for hardware produced by a long dead company...

PCI/ISA Modems...
They are kind of standardized...unless you're talking about a softmodem of some sort...which you should steer away from anyhow...

NICs...3com...

However, I suggest RealTek based cards because they are so cheap in comparison and seem to have good support.

Sound...Creative...

Re:I second this...

[Zathrus]

I think making the specs public is all we ever really asked for.

Which is akin, in some cases, to saying "come on in and take the kitchen sink while you're at it" for hardware manufacturers.

The Linux community (and the OSS community at large) needs to get over this. Open Source is fine and grand, but it's not always viable. With that in mind, a company should either make the interface available, or make reliable, fast, and solid drivers available on a regular basis.

Those that choose neither may very well be reviled. Those that choose one or the other should be praised. And those that choose to reveal the interface AND help in writing the drivers should be revered.

But bitching about a company that chooses to keep trade secrets secret is really f'ing stupid.

Linux Media Labs

[AJWM]

This company designs and manufactures its own video capture and compression cards, and also remarkets some third-party cards. All supported under Linux (with a name like that, what else?) and they GPL the drivers.

Pretty cool.

Re:Typical response

[SailorMeeko]

I've got to give a big thumbs up for Earthlink tech support.

About a year and a half ago when my brother got his cable modem, I helped him setup a Linux server (Red Hat 6.2) + IP Masquerading for his home network. This was fairly straightforward, and required no tech support help, and has run fine ever since.

Well, this weekend he switched to Earthlink DSL. I wanted to make the switch as easy as possible for him, without having to re-install anything. Unfortunately, I could not get it to work as easily as the cable modem had, and so we called Earthlink tech support. After trying to ask for help without revealing which OS we were using, it was becoming quite obvious we couldn't. The person asked what OS we are using, and we told him Red Hat Linux 6.2 We were expecting to hear something like "We're sorry, that is not supported", but instead he told us we need a PPPoE in order to connect. He then told us about a program called Roaring Penguin, and where to go to download it. He then helped us configure it and get us connected. We were both quite impressed.

I also remember a time about 4 or 5 years ago when I had Earthlink dial-up, and I was using something like Red Hat 5.0 or maybe even 4.x Anyway, it was in the earlier days before establishing an Internet connection in Linux was easy (at least for me), and I was having trouble getting connected (editing ppp scripts and such, all from the CL), so just for fun I decided to give Earthlink tech support a call. To my surprise, the person walked me through the editing of the scripts, and they worked perfectly, and I was connected.

Re:I second this...

[ImaLamer]

I second this!

I praise the points the OSS community makes when it complains about people like nVidia, who release drivers but not trade secrets, or a company which only gives the community so much information to get the device to work.

I understand their need. What if it doesn't work. Not having the source makes it almost impossible for there to be a good work around. What if you want to do something cool with your device? You can't until more is known.

BUT! In some cases your device manufacturer may have just sold that same source code for a pretty penny. You can't expect them to publish on the web as well. If they are subcontracting - making hardware for a non-computer company then don't expect them to reveal all the secrets.

Let them support the hardware first. Sometimes there are things you may not understand that keeps them from just saying:

"HERE IS HOW THE DEVICE WORKS! - COMPETITORS PLEASE COPY IT"

The hardware business is one of secrecy.

NVidia's closed source drivers cause problems

[FreeUser]

NVIDIA actively support Linux by constantly releasing up to date drivers that are very high quality. The NVIDIA drivers are unquestionably the highest quality OpenGL implementation available on Linux without exception.

Well, perhaps ... but NVIDIA's closed source drivers, while good in some respects, do occasionally cause X to hang for no apparant reason. Switching NVIDIA cards, or updating to the current drivers, does nothing to alleviate this, although switching from an NVIDIA card to an ATI Radeon card did solve the problem, as did using the Free Software Nvidia X driver ('nv') with the same hardware that was so troublesome with the 'nvidia' driver. And yes, this is with AGP settings in the safest, most conservative mode (cf the NVIDIA driver docs for details).

So while the OpenGL implimentation may be very good, the closed source nature of the driver means I'm forced to wait for an officially unsupported, binary-only driver, to be fixed someday, or I have to find an alternative. This seriously decreases the value of the NVIDIA driver and hardware for use where I work and live.

ATI does not suffer from this handicap, and while its OpenGL support may not be as good as NVIDIAs, it does work well, and without the system stability issues incurred by using NVIDIA. In addition, the free and open nature of the ati drivers insures that my hardware will never be orphaned, even if ATI has a change of heart (or financial troubles) down the road. The closed source NVIDIA drivers give me none of those guarantees (though the fallback nv driver helps, as long as you don't need digital out or multi-head support).

Re:NVidia's closed source drivers cause problems

[vidarh]

I'd second this. My next graphics card will most certainly NOT be an NVIDIA based one. I can live with somewhat lower performance, but not with having my machine crashing all the time.

Re:NVidia's closed source drivers cause problems

[Afrosheen]

You must have some hardware issues that are agp related or something else. I've had a geforce2 mx for months now and have used all of the nvidia drivers on mandrake (even through 4 beta releases of 8.2) and never had ONE freeze. Don't spread FUD based on your personal experience with shite/buggered hardware.

Re:NVidia's closed source drivers cause problems

[T3kno]

Amen to this, I've been using nvidia cards for years, and have not had that many problems with them. KDE freezes my machine more than the nvidia drivers (properly configured of course).

Re:NVidia's closed source drivers cause problems

[dinivin]

Believe it or not, moron, a lot of people have had lockup problems with nVidia's drivers.

Just because you haven't had those problems doesn't mean that what he posted was FUD.

Dinivin

Re:NVidia's closed source drivers cause problems

[dinivin]

I called him a moron because he deserves it. He automatically accuses the original poster of havning hardware issues "that are agp related or something else".

Why is it that so many of you nVidia fan boys refuse to accept the fact that on some pretty common hardware, the nVidia drivers still have problems for some people?!?

Dinivin

Re:NVidia's closed source drivers cause problems

[Dwonis]

Don't get a G550 (or probably a G450, for that matter) -- the drivers are much crappier (from my perspective, they weren't too much better than nVidia's drivers).

Re:NVidia's closed source drivers cause problems

[FreeUser]

hy is it that so many of you nVidia fan boys refuse to accept the fact that on some pretty common hardware, the nVidia drivers still have problems for some people?!?

Actually, the binary NVidia drivers cause problems on a lot of systems. I have 20-odd systems at work with various NVidia drivers and hardware, running X in various configurations (some dual headed, some single headed, some PCI cards, some AGP, various processors, various versions of X, etc.).

They all suffer from these kind of stability issues, sufficiently often that those which are not using the dual-head features of the cards have been switched to using the free 'nv' driver, which lacks many of the features of the closed-source drivers but are significantly more stable and reliable.

For the rest, we are no longer purchasing NVidia cards because of the stability issues, and are migrating toward ATI and other solutions. Binary-only, proprietary drivers simply entail unnacceptable risks in stability, and leave our hands tied when problems arise.

The NVidia fans can complain, accuse myself and my predicessors of incompetence (they were not incompetent and neither am I, but that is neither here nor there), claim there are 'AGP' issues even on PCI-only systems, etc. etc., but when all is said and done, removing NVidia's binary-only drivers and replacing them with the free 'nv' drivers, or replacing the hardware with ATI hardware and its respective free drivers, turns the machine from a flakey box reminisicent of Windows stability to a rock solid, reliable workhorse with uptimes measured in the hundreds of days.

There should be a "Linux Friendly" award...

[maroberts]

..with varying grades to hardware manufacturers, combined with a logo that can be placed on packaging. Say a Linux Friendly logo, with awards for a product ranging from bronze to platinum, depending on how much the manufacturer supports Linux.

Lexmark

[vidarh]

I thought I'd never say this, but Lexmark actually seems to be getting quite Linux friendly. They specifically list Linux on several of their websites for a reasonable subset of their printers, and have drivers available for downloads...

It's not more than a year or two ago that I had to return a Lexmark printer because I couldn't get it to work with Linux at all.

Great ones:

[Kether]

Syskonnect - Makes some of the best network cards, and developes very, very stable drivers for linux. Best of Breed.

ICP-Vortex - Makes some of the best RAID cards available. Develops their own drivers. Best of Breed again.

Cyclades - makes some nice stuff, supports linux well.

Adaptec may have gotten better, but they didnt used to release the source for their RAID cards, and only realeased binary drivers for certain kernels, and didnt update them often.

Mylex used to advertize heavy about working with linux, but relied upon community drivers, even linking to the community page. Why woudlnt they bring this person on-board to fully support linux?
Makes no sence to me. Why buy Mylex when i can have a much better card in ICP-Vortex anyway?

Nvidia rocks, ATI sucks... butdon't forget Matrox!

[EvilAlien]

License, schmicense. Who cares... Nvidia rocks, they release timely updates to the drivers, their cards take the performance and quality crown, and they seem to genuinely care about supporting Linux. I wish the same could be said for ATI, who may be cooperative with the community at times, but not consistently. I have nothing but bad things to say about ATI, and I'm embarassed to have ever purchased their products.

Another hardware vendor worthy of note is Matrox. They've done a good job in supporting Linux, and deserve recognition for that effort.

NOT KODAK

[cowbird]

Kodak is one of the real losers when it comes to releasing specs. Scratch them from the list, and add them to your blacklist.

Linksys...

[gillbates]

I have a few Linksys ethernet cards which came with Linux drivers - in binary form, though. I have started to see Linux mentioned on more and more hardware boxes, and this is a great improvement from printing out RedHat's hardware compatibility list and then going shopping. Plus, my mom can now by hardware that will work with our Linux systems, and I don't have worry so much about her buying junk that will only work with windows....

Re:"Secret technology assets" are mostly illusion

[CDWert]

Well....Here goes one last time....

The fact, however, is that that's mostly bullshit. For 95 (maybe more like 99.5) percent of the ideas out there, a competent designer"

I agree 100% with that, but the problem isnt the 99% its the 1% that they are afraid and rightfully so, of releasing.

"As for patents, they are allowed to release those details. In fact, the whole point of a patent is to give a monopoly in exchange for public disclosure of how something works."

True BUT, my point was actully the IMPLEMENTAION of theose patents, some of which are covered as Trade Secrets, and there are NDA's to given implementations, a patent doesent have to give you the process leading to the implmentation of something. You can patent a chemical WITHOUT releasing the manufacture method.

I guess the singly most important statement you made is "It's true that small market segments are often ignored, and justifiably so, but that does't mean that the people in those segments aren't entitled to notice that the support they get is substandard."

The truth is FOR THIS GIVEN MARKET SEGMENT, their support is not substandard, it may in your eyes be substandard to windows support, but as 3d accells on Linux go its far from substandard. Dont even bark about the ATI stuff, its crap up toll the last gen and they dont release ALL of the details of what they have released in the last 6 months in competition with Nvidia.

Standard and Substandard are RELATIVE terms,I think in a area(hardware) where most compaies dont support Linux to any Extent Nvidias support is far from substandard, if you comparing Linux to Windows support I would agree. But then again apples and oranges.....

The Role of Hardware Design

[Chris Tyler]

Beside releasing specs, the other thing that the hardware companies can do is support standards and engineer their products so that there is less variation between product generations.

Rather than invent new protocols, command sequences, and interfaces, they can support a standard interface across their whole product line.

This makes it easier for the open-source developers, but it also makes it easier for the company itself -- hardware designers, in-house developers, and support people. In many cases, an old driver can be used, perhaps slightly updated to manage a few new features. This reduces the amount of redevelopment and therefore reduces the opportunities for bugs to sneak in -- regardless of the platform.

Some good examples come to mind:

- HP scanners. The HP scanner protocol has been pretty much stable for years, and the same command set has been used on the USB scanners as the SCSI scanners. You can take a current SCSI scanner and use it with a driver from 6 years ago. Yes, the protocol is proprietary, but it's well documented and well understood, and it's not changed at whim.

- DPT controllers (old). These used the EATA (extended ATA) interface across the product line. EATA was well-documented, multi-vendor, and stable. It provided basic compatability with ATA (IDE host adapter) specs but could then take off from there. New cards needed tweaking but not wholesale driver rewrites.

- Most SCSI tape drives. These all use the standard SCSI tape command set, even though they have very different capabilities. (Contrast this to OnStream drives, below).

Some bad examples:

- Early OnStream tape drives. Although the newer units understand standard SCSI tape protocols, the early units used an unnecessary proprietary variation. There were reasons for the variation -- but the fact that the newer drives understand the standard command sets indicates that the variation was not necessary.

- Video cards. Why can't successive video cards from the same manufacturer each support a superset of the previous capabilities, so that you could use the previous driver to start, then eventually add the new functionality to the driver to fully support the latest card?

- Many advanced laser printers (this is a cross-manufacturer issue). I have yet to see two different makers that use the same paper-source-select or staple-enable codes. If PCL and PostScript and PJL are all standardized for other functions, why not source-select and finisher options? It wouldn't require an ANSI subcommittee, just one or two face-to-face meetings or a couple of days of faxes and e-mails.

In most cases, these are engineering problems. The first-generation products need to be designed with some foresight -- version numbers, capability registers, extensible command sets, protocols that can be implemented over different interfaces -- so that later product generations can interoperate, even when they support features which we can't even dream about now.

-Chris Tyler

Re:Nvidia rocks, ATI sucks... butdon't forget Matr

[Afrosheen]

All I have to say is this:

Buy a radeon 8500 and try playing quake3 or anything else 3d in linux. Impossible.

Buy an Nvidia Geforce whatever or a lowly tnt. You'll be playing in 5 minutes or less.

Which one would you buy? It's obvious. Even though ATI provides a little community support, they don't personally work on the drivers and release them. Closed source, open source, the point is that ATI support SUCKS compared to Nvidia. All the DRI people say Radeon 8500 3d support sometime but it's taking a long, long time. I'll stick with Nvidia, a company I know and trust and can count on to supply me with high quality drivers.

3Com

[King_TJ]

Yep - I agree too, except 3Com network cards aren't always all they're cracked up to be.

It's not so much a problem with driver support (or lack thereof), but they've produced millions of cards with flawed chipsets/hardware.

Ever use/see a 3C595 10/100 PCI card, for example? They've got issues. I've had a number of auto-sensing switches that wouldn't work with these cards unless you locked the cards at 10 mbit. first. Even on a cheap D-Link 8 port switch at home, I only get half-duplex operation whenever I plug in a 3C595.

There are acknowledgements that these cards were buggy, but you really have to dig around to find it stated on the manufacturer web site.

Then too, you have the extremely popular 3C905 (Etherlink XL) series adapters. Good cards, but 3Com made numerous revisions to them over the years - causing lots of confusion. Their latest Windows drivers simply probe the card and automatically deal with whichever variation it happens to be, hiding the problem from the end user -- but it's a hassle for others. (Look at www.rom-o-matic.net, where you can download Etherboot disks, and see how many 3C905 images they have posted up there!)

In my opinion, a revision that renders the previous driver completely unusable deserves giving the card a whole new model number.

Re:Any company....

[Afrosheen]

Idiocy didn't create winmodems, bean-counters at major companies did.

Marketing guy to Engineer: "So let me get this straight. If you guys make a huge, processor intensive driver that takes over the function of the dsp chip on current modems, then we can cut our costs by 4/5ths?"

Engineer: "Yes, but it'll suck and the drivers will only work on windows because that's all our programmers know."

Marketing guy: "Doesn't matter. Everyone uses windows anyway.We can market it as a win-modem for a lower cost!"

Marketing guy to bean counter: "We can make new modems for 4/5th the price of current modems with these new drivers! We can either slash prices on the new stuff or increase our profit margin to ridiculous new heights!"

Bean counter: "Excellent! Tell the CEO right away!"

And thus it began...

Axis Communication

[KjetilK]

High on my list is Axis Communication. They have a nice chip with lots of neat stuff that runs Linux, and they also got this chip in printservers, bluetooth access points, etc. I think those are they guys who wrote most of the Bluetooth code in Linux.

I really want a BT access point, I was just hoping the prize would come down a bit to make it viable for home use. I really don't need the range either.

EDT

[spagthorpe]

The guys at EDT (www.edt.com) who make all sorts of data acquisition and control cards, support Linux very well. I've used their LVDS cards for a handful of projects, and they are very knowledgeable and helpful with Linux. They have native drivers for the cards, sample code, etc. Highly recommended. I have no connection with them other than being a very satisfied customer.

Keyspan usb/serial adapters

[beefubermensch]

Keyspan supports Linux. We're about to release a new version of the driver, which we've modified in-house to fix many bugs. I wouldn't say we excel at Linux, but we're interested in it, and as far as I know our policy is to devote as much time as we can based on estimated sales into the Linux market. Anybody out there trying to run a headless server on a machine without native DB-9 ports?

Dunno if it's "active" support, but...

[seebs]

When I wanted to get drivers written for a Linksys USB->802.11b adapter, they didn't *donate* drivers, but they offered to sell me a couple of units at their "employee discount" rate for development and testing. I'd call that fairly supportive.

VA?

[Mignon]

Ok, well I guess VA doesn't make hardware anymore, but I recall they were active in writing RAID code for the hardware they shipped. Of course they also have (or had, I don't keep up with this) quite a few heavy-hitter kernel developers on their payroll.

Re:Matrox

[Arandir]

My G450+ works just fine with 4x AGP and OpenGL acceleration (without even having to use drm-kmod).

Re:I second this... - Well, I don't

[Nurf]

Which is akin, in some cases, to saying "come on in and take the kitchen sink while you're at it" for hardware manufacturers.

As a hardware designer I can say this is almost never the case. If you're any good, you know exactly how something was done. This doesn't mean that you can program the hardware though, because there are still too many ways they could lay out the interfaces.

I'll say it again: It is extremely rare that you are giving anything "secret" away by telling somehow to interface to your device. I'm sure lawyers like forcing their engineers to keep quiet, just in case, but there is no real technical reason to do this. It's akin to designing a car and using your own controls, then not telling drivers which pedals do what.

I see fear here, not reason.

Re:Matrox

[Arandir]

My Matrox G450+ works just fine under FreeBSD-4.5, using a stock XFree86-4.2.0. I get kernel accelerated OpenGL and the best 2D acceleration I have ever seen off of any card. And it worked out of the box! No drm-kmod or hal needed. No bizzare kernel configurations or init scripts. I just needed to copy the kernel modules out of /usr/X11R6/lib/modules/dri and into /modules. This is on an A7M266 mobo.

Re:I second this...

[anshil]

I don't agree at all

When I'm a company there is not real technical problem in buying a competitor product, open it and look how it functions. It think it's legelly even legal. Don't come me what you thing is "right", but as I understood the laws reverse engineering is _not_ illegal at all. (as long there is no "valid" patent on the used technology)

I use and used to be working with embedded products, and everytime I get some interesting box in my fingers (i.e. a friend buyed a firewall, or a modem, or a fax. I find it frilling to open the box, to take a look what CPU they used, what RAM, ROM or Flash they could get/buy, how the print is layered, how much layers they needed etc.

Secondly when I buy a hardware product I buy exactly this, a hardware product, and not a "functionality extension within an os". So I feel it's my right to interact with the hardware, and to know how my PC interacts with it. The company may not do on the PCI bus like it wants, it also has to keep the voltages in valid ranges and all that :o)

In general why I guess that some companies don't give out specs for the interfaces is because the reponsible posts just have no spines. Will it cost my job _not_ to everything secret? guess no, can it cost my job to publish something? If you publish the wrong thing yes, However if you publish the right thing it can get you a raise. For me it's a matter of spine to make career.

From another standpoint, people are always thinking whats best for companies? Thats not the right way to think as a human. A human should ask himself whats best for the society as whole? I mean every company has to watch to keep alive, thats what it should do, but never should forget what companies are for after all. To add something for the society, the community of all people. And there a lot of the greedy all-to-me philosophy fails. But in the long run a freer community will survive eitherway better, as people help each other out, show them the interfaces to their boxes etc. while a parallell closed community will just survive less good. Thats how evolution _will_ take place. It's just a question, do you want to be part of a closed no-one-communicates-with-no-one-just-in-case-commu ntiy or and community targeted to work a bit together. I guess one can estimate witch one will survive in the long run. (some decades, maybe centriues). At least I want to be on a living branch of society, not the dead one.

Sorry, I can't agree

[HiThere]

If you get a binary dirver, you're getting something that will work until the next time you upgrade your system. If you get the source, then you're getting something that will work until you change your hardware.

I upgrade my OS frequently (perhaps excessively). So binary drivers aren't worth that much to me. And this is *WITHIN* the x86 hardware line. Now if I wanted to try a different cpu ... well, I might need to edit the source. But just forget porting the binary!

Binary is a nice accompaniment to the source, but it's not a replacement. And I'm not even a driver developer. I depend on other people to have put together the drivers that I use. But I still need to be able to adapt for when, say, I change libraries.

N.B.: This is also true for programs of other sorts. The only binary file's I'll buy these days are games. And that's because it doesn't hurt me if they stop working. Which they do.

Software people have much better reasons to be upset about this than hardware people do. If you're selling me a dololly that plings the inghams, why should you care which version of the OS I use? Your aren't giving away any great secrets by telling me which pins to signal when I want to activate it. The software people generally either need to provide their entire source code, or statically link the entire OS. They have reasons to be upset by this. But they don't have the right to take my money for something that stops working immediately. If I have the source, I can probably recompile it, and I might be able to fix it (if not, someone else can). So it's worth my while. If I don't, I've thrown my money down the drain. And if that's what they're asking me to do, we've got a problem:
They don't have a customer, and I don't have a vendor.
But better that than that I've wasted my money on something worthless.

I'm sure that there are other answers. It seems to me that they should be able to build "compatibility libraries" that can both be compiled on whatever the current system is and can be called by their program to do it's work. But so far nobody seems to do that (except some of the open source projects). Now I will grant you that this probably wouldn't suffice to overcome major version upgrades. The kind of which they say "This will break binary compatibility". At least not usually. But that they could generally overcome by just recompiling the internal code (which I will grant they have a reason to keep secret, if they grant that I have a reason to insist on it working through upgrades).

3ware

[Paul Johnson]

They write and then open-source their Linux drivers.

I haven't actually used their IDE-RAID cards, but everyone I've heard from speaks very highly of them.

Paul.

Re:Just which drivers do you _have_ to compile in,

[Wavicle]

I'm not talking from a completely baseless pedestal here. I've written a Linux device driver, and am well aware of how loadable modules work. I am also well aware that everybody who tries or has tried distributing a binary module has had trouble. In fact to the best of my knowledge, nobody has built my driver from source (separately from the kernel source) as a module that could be loaded into the mdksecure kernel - so no binary would work for that anyway (and if you're using mandrake, there is a high probability you'll end up using that kernel).

How many binary module configurations would I have to make available to catch a reasonable swath of Linux users? How much would it cost me in support everytime someone says "Hey, when I add your module it says I might taint the kernel"?

What reasons do they have for not releasing specs? Well, in the case of my driver, the manufacturer (a fairly small player in the imaging industry) had come up with a clever way of doing color images at 8 bits per pixel with more color information than any other 8 bit per pixel color scheme. This means their camera can send either more fps than a similar quality camera, or more quality than another 8 bpp camera.

There are a lot of small players who have come up with something clever and would rather keep it as intellectual property than patent it for the world to see. The majority of users don't care if its open source or proprietary, they just want it to work.

DISC Inc.

[Eric Green]

DISC Storage.

Not only do they have all the SCSI specs for their jukeboxes and optical drives online so driver writers can easily access them, but they pay my salary too :-). Right now all DISC/NSM hardware is supported by the 'mtx' media changer program for Linux (which I maintain, which is distributed with Debian, RedHat, SuSE, and probably other Linux distributions). The only thing that does not currently work is importing media via the import/export slot on the NSM DVD-RAM libraries, which because of the hardware involved needs extra support (the standard way of handling import/export -- send a MOVE_MEDIUM command with a source and destination address of the mail slot to tell the jukebox to stick out its tongue -- doesn't work because the hardware must know which slot you're going to import the disk into before it sticks out its tongue, because it must move that caddy to the mail slot -- the caddies stay inside the box, you get the bare disks spit out at you or you insert bare disks rather than have to mess with caddies yourself). I'm currently working on adding support for those special features (and features such as the disk pack mechanism) to the 'mtx' suite.

-E

Why binary only is bad

[jmorris42]

If you need to update your kernel to get the latest shiny top working and your binary only vendor has not got around to releasing a new version for that kernel... you are screwed.

When said vendor decides your product is 'end of life' and you want to apply a new kernel to close a security hole, you are screwed.

I could fill pages with variations on the theme but anyone who hasn't got the point yet won't.

Sangoma!

[broody]

The vendor whose Linux support that simply blew me away was Sangoma! They wrote a quad port PPP over HDLC Linux driver for the S5142 WANPipe card to solve my company's problem. We needed a router to handle 8 PPP ports for an old leased line and thanks to Sangoma we had a cheap and robust Linux solution.

That was back on 2.2.16. I remember some changes in the PPP kernel code caused problems on releases above that but I am sure it's ancient history now.

I cannot recommend Sangoma highly enough. Great product, service, and people.

Closed specs

[jmorris42]

No, actually I come from the old school where all hardware came with programmer's information. It was just an accepted and expected part of buying hardware. Then the 'end users' came along and companies stopped providing it in the manual to save space but would provide it if you asked. Somewhere along the line they decided that information was now a vital trade secret.

Bull. If your hardware is so lame that letting anyone see how it works would destroy it's value it probably didn't have much to start with. I know hardware reviews would be a lot more informative if real information was still available.

Since NVIDIA is the popular whipping boy today, lets use them as an example. Assume that the popular belief is true and that much of the value of their hardware is in their drivers. Open sourcing them would give away valuable secrets so they might not want to do it. Fine. Details on the interface between the software and the hardware still should not be harmful to their secrets. If their drivers really represent most of their value it might be a long time before the XFree nv driver equaled theirs, but that would be ok by me. I actually use the closed driver with an old TNT2, but I'd feel a lot better about buying a current card if I knew the investment was safe.

A cheap way for HW vendors to support Linux...

[Ungrounded Lightning]

ATI supply the specs but apart from that do almost nothing, they have enough trouble supporting high quality Windows drivers. The reason you need the specs is to get any kind of driver support at all, when the manufacturer is delivering full high quality up to date drivers with more OpenGL support and extension support and quality than anyone else I'd rather have that than specs and a driver development effort that can't keep up.

Here's how hardware developers (and not just ATI) can get solid Linux support on the cheap:

If:

You haven't done a Linux driver yet.

You have done a good Windows driver.

You OWN the source code for your Windows driver.

The source code doesn't leak a deep dark trade secret (if it does - PATENT it and then it won't).

Then release:

The Windows driver source under an Open Source license, along with...

documentation of the device. (That's typically schematics, chip specs, and maybe some internal docs and/or memos from the development team.)

You already have it. Vet it for any deep secrets and licencing problems with your partners, but otherwise don't bother to clean it up. Just dump it on us.

I'm sure that if your device is AT ALL interesting somebody in the Linux community will be GLAD to port your driver - and any future upgrades.

Re:You fucking moron

[Ogerman]

What about drivers for Windows? I have seen very few drivers for Windows that WEREN'T closed-source...

What does this have to do with Windows? I don't use Windows. NVidia's binary drivers are a pain because they have to try to keep up with every change made to X and the linux kernel. Frankly, I don't know why they'd want to do it themselves and not just give out the docs so the community can develop them.

Oh, and by the way, the open drivers for ATI cards work beautifully. Dunno about Windows, but who cares. Windows sucks anyhow.

Cost of changes

[Dwonis]

Yep. It basically comes down to the cost of getting the changes YOU need in the drivers. With open-source drivers and a good set of specifications, if your drivers are missing some functionality or you need a really annoying bug fixed, you can:

• Pay a programmer to fix the problem. Assuming that programmer makes $60/hour, and the changes will take 30 hours, the cost to you is approximately $1800.
• Fix the problem yourself, if you are able. Only costs your time.
• Ask someone else to fix the problem with your help. For example, making a polite request on the mailing list for someone to help you. Many open-source developers will help you with the problem, especially if you buy/lend them the card you're using, which would cost no more than $400 (and that's an expensive card).
• Wait until someone else encounters the same problem and gets it fixed using the methods above. This is a risky option, since it may never happen, but it still does happen quite often for widespread bugs.

If nVidia doesn't think it's worth it to fix the problem, you can:

• Nag nVidia, talk to a salesman, try hard to find the right person to call, call that person, find out that you have to pay a large sum of money to licence nVidia's source code (which you may not have -- most people stop at this point, if not sooner), then use the methods above to solve the problem, then figure out how to get nVidia to roll the changes back into their version. You can estimate the cost of that.

So, given nVidia's model and the open-source model, which would you take?

Software commoditizing

[Dwonis]

I mentioned it in another thread, but I'll write it again:

Software -- especially operating systems -- is becoming a commodity. Companies that maintain "90% of users only want..." or "Best viewed with..." policies will not survive.