A constant thorn in the eye of many Linux kernel developers is the existence of closed-source kernel modules, most notably those by Nvidia and Ati, but also some file system drivers and other elements. Most of the Linux developers have been against these modules ever since they were first used, and in fact, bug reports originating from a tainted kernel are often disregarded and ignored. The kernel developers have now rallied together by issuing a statement urging vendors to release open source Linux kernel modules and drivers.The statement is fairly clear, and provides little new information to people even remotely familiar with the Linux world. It reads:
We, the undersigned Linux kernel developers, consider any closed-source Linux kernel module or driver to be harmful and undesirable. We have repeatedly found them to be detrimental to Linux users, businesses, and the greater Linux ecosystem. Such modules negate the openness, stability, flexibility, and maintainability of the Linux development model and shut their users off from the expertise of the Linux community. Vendors that provide closed-source kernel modules force their customers to give up key Linux advantages or choose new vendors. Therefore, in order to take full advantage of the cost savings and shared support benefits open source has to offer, we urge vendors to adopt a policy of supporting their customers on Linux with open-source kernel code.
The statement has been signed so far by more than 135 kernel developers. The FAQ accompanying the statement explains that “nothing has changed, we have just been receiving a constant stream of questions from companies asking how the Linux kernel developers feel about closed source modules over the past year or so. This statement should be the definite answer for how a large majority of them feel with regards to this topic.”
The statement will most likely do little to force vendors to write open source kernel modules, since effectively, this statement changes nothing about the situation as it was – it just sort of formalises the whole thing. We already knew the kernel developers don’t like closed source modules and drivers, and we already knew that some vendors simply don’t care about being open source. They just want their drivers to work, and seeing many end users are just fine with using these closed source modules, vendors have little incentive to change anything about it.
I find that a sad thing, but it is the way it is. I’m too practical to give up my closed source Nvidia driver, and in addition, it would be quite hypocritical too [just to note: I’m referring to myself here, not anyone else!]. I can be all noble by disabling closed source modules and even software from my Linux install, but then I’d go to bed at night and load up my PowerBook running Apple’s Mac OS X to check the web just before I go to bed. I’d wake up the next morning, and boot into Windows Vista to work on university projects using Microsoft Office, while I check my Windows Mobile phone for text messages.
It would be like being a part-time vegetarian. I’m a part-time vegetarian too; I never eat meat, except at 6pm when it’s dinnertime.
Perhaps I’m being a dunderhead but where’s Linus ?
Working on the kernel instead of pushing their agendas in attempts to oppress peoples software freedom?
GPL isn’t freedom as it pushes everyone to submit to one agenda.
And when was the last time you were pushed to use GPL software, or pushed to link GPL libraries?
some people.
Would FOSS kernel modules for wifi networking (cough, broadcom, caugh) and other hardware only supported now though binary blob “we have patents in there” bad coding really be such a bad thing.
As far as I can see, the only limitation is that everybody must be allowed the four freedoms. There are many other licenses though so if you don’t like GPL, choose something different for your code.
(Are you a developer who would actually be effected by the “limitations”? I can write code but I wouldn’t claim to be a developer or limited by GPL.)
(Edit: drat, that was meant for the complaining post above yours)
Edited 2008-06-24 13:50 UTC
You can choose not to use gpl software. This is freedom. However the fact that I have to use windows in order to make my printer function properly is definitely not freedom. When I bought it I didn’t know that it was win-only. The next printer was linux-oss friendly. But who will give me back the money for my old printer? Worse is the fact that it has no Vista drivers.
So mister troll will you give me back my MONEY??????? If the specs were open I could write a CUPS driver. I know how and I would do it.
When you buy a printer, you first check the requirements first, if is not designed or intented to work in linux try another printer, simple. But not try to make it work in linux if is not intented for it.
You vote with your money,
If that “voting with your money” idea worked, we would be living in a completely different world.
The sad truth is that the majority don’t care.
Haven’t the majority then given a vote with their money? In case you missed, the answer was: “We don’t care!” (or don’t known or are too stupid to realize or whatever)
If I design a software product and people say I cant make it play with GPL product X because I wont hand over my hard work to others, that’s not freedom.
The designer of gpl product X can refuse to take my closed source product as part of his product, that’s his freedom…but he cant tell me I cant make a closed source product to work WITH his GPL product.
Closed source isn’t going away…people like to hold their hard work in their own hands way too much…so best thing for GPL people is to play nice with the closed source.
The harder GPL people try to fight closed source, the more they are shooting themselves to the foot.
The fact there IS Nvidia and Ati drivers for Linux is a victory for linux by itself…telling Nvidia and AMD to hand over their work or be gone is pure selfishness and inflicts THEIR freedom.
You have the freedom of choice to use or not to use GPL software; nobody is forcing you. Why do you think you have the right to disrespect the freedom and wishes of the GPL software authors who GIVE you their work with very modest conditions?
All a GPL author can do is demand that you conform to the legalities of the GPL license. He can also hope against hope that you are honourable enough to respect the intent of the GPL and not try to skirt around it with a legal loophole.
GPL software isn’t going away… its authors like to put their hard work in the hands of people all over the world…so the best thing for non-GPL people is to respect the modest constraints of the GPL code if they wish to use it.
No, you’re just plain wrong. The folks dedicated to creating a healthy and vibrant open source eco-system with GPL licensed software are not served at all by looking-the-other-way when closed-source providers violate the spirit if not the legality of the GPL.
You have this exactly backwards. It’s companies like Nvidia who are circumventing the freedom of the GPL software authors to have the license of THEIR choice respected.
Edited 2008-06-25 00:51 UTC
This has nothing to do with FSF, so please stop your trolling. This has everything to do with the kernel development team wanting to see the code for modules that are being inserted into the running Linux kernel.
I personally find the linux kernel developers request offensive.
Dave
He’s been very vocal in the past if I read correctly. Maybe he’s already made his point and is just sitting back to see how the other’s decide. He is the posterboy and original coder for the kernel; this may be between the other Dev’s now.
I suspect we’ll hear something eventually.
What is it with everything having to be open source? These are drivers for commercial hardware and the manufacturers want to protect their IP. Yes we’d all love it to be open source but if the manufacturer cannot allow that then fair enough, I’d prefer that than no driver or a reverse engineered one.
We have to let open source and closed source play together.
If they cannot allow people to use the products they bought and paid for, maybe they should not be in the business of selling hardware?
For linux, the proper place for drivers is in the kernel tree.
That is not about politics or religion, but practicality. Otherwise the users will have to wait around for support for their new distro/kernel/magic beans to not only arrive, but also mature.
The manufacturers do not even need to open source their work – just provide proper specs for the hardware and someone else would do the work for them!
There is a long way to go, but for (linux) usability, it is probably the best option.
I guess I would trust NVidia to insert a driver in my kernel but I admit I wouldn’t want to be doing this for all my devices.
I don’t think its practical to bring out a new piece of hardware and wait until someone reads the specs and writes a driver – of course it would be a happy medium to provide both. Bring out a closed source driver and the hardware specs, so you can use the closed source driver now or wait until someone (or you) writes an open source driver.
It’d be even better of the vendor started the driver under a FOSS licence then worked with or handed off development to the Kernel devs, the Linux Driver project or any other interested developers.
The end user would gain the best of both worlds; driver available at hardware release and ongoing maturity.
I have another option for you. Pay for the closed driver until someone reads the specs for you to use it. Maybe then you will show respect to standards like VESA/FIREWIRE/BLUETOOTH and so on.
Do you have the sourcecode for the engine computer in your auto? No ?!?! You should ask the manufacturer. By golly you should be able to examine and recompile your engine computer code. Maybe you want to recompile the braking system computer code too..
This is one of the few places the car analogy breaks down when discussing computers and software. I know where the break line is. I know where the break pads and plates are. I know where the stearing column routes through the engine block. I don’t touch either of them. I’ll change my oil, tinker with the engine and all sort of other things but breaks and steering are for the local certified mechanic.
If I recompile my kernel; I don’t risk hitting a tree if it crashes. And, with the bad ram I just pulled out of my rig, I’m very glad for that with the number of crashes those Balistix caused.
It’s also about practicality, not religion as a post above clearly points out.
The kernel is the chunk of code between the hardware and everything else. The kernel’s whole job in life is to recieve requests from the rest of the system, translate it, then forward it on to the applicable hardware. It then takes the hardware response, translates it and forwards it back up too the rest of the system code. (A simplistic description I know.) This is why known hardware “just works” on Linux based OS, the BSDs or other truly Unix like systems without aftermarket addons.
The natural and most practical place for hardware support is within the OS kernel not as aftermarket addons.
Now, what’s better for the end user?
– having to install the hardware then install the drivers plus whatever crappity software is forced on them. (I’m looking at you ATI media payer.)
– having to install the hardware and reboot.
“- having to install the hardware then install the drivers plus whatever crappity software is forced on them. (I’m looking at you ATI media payer.)”
Just download the driver only. I certainly don’t have an ATI media player installed on any of my machines, and they are all ATI.
in the case of ATI thats true. I ended up doing that leaving me the challenge of finding tuner software and dvd codecs. this was a while back and AMD seems to be turning it around. now I’ve an nVidia, hauppauge and striker2 motherboard I’m fighting with under Windows. woks under other OS except the hauppauge and full 5.1 sound; mostly due to limited time.
Actually, you CAN get an open source engine computer, it’s called the MegaSquirt (http://en.wikipedia.org/wiki/MegaSquirt) and it’s supposed to be pretty universal for compatibility. Yes, there is an advantage over the proprietary factory computer in that you have full control over all the fuel/air mixture settings and what not which you are normally locked out from messing with. Otherwise you’d have to pay $500 or more for a proprietary performance chip as opposed to $200 for this thing. You do however have to setup the whole system yourself, kind of like Linux From Scratch, but there are third parties that sell these things already setup for your car.
GCrain. Bad example.
With your CAR, do you have it hooked up to the internet? Do you have multiple programs running that you wish to customize and make sure it works together? Do you CARE about security holes in your cars software? Why would you? There’s not an issue there. However with binary blobs, they can cause instability with other pieces of software, etc – something you WON’T expect to happen on your cars ‘software’
If a piece of hardware cannot be used, for whatever reason, in the environment you intend to use it, then maybe you shouldn’t be buying it. If a company is happy with their hardware only being bought and used by Windows users and they’re making a profit on the business model I really fail to see the problem.
It’s not that I’m not an OSS supporter, I am. I also think that it would be great if more companies would use open source for their products. But I don’t think they should be forced in anyway. They should do so because it’s good idea for their company. If they think they can make more money and run their business better using close source then I’m fine with that. I may simply chose to not buy their product.
Buying hardware that has no or bad Linux support and then complaining that it doesn’t work under Linux just doesn’t make sense. It’s not like these things are impossible to check out before buying.
If I need a graphics card with DirectX10 support I make sure it has it before I buy it. If I need a graphics card with support for a certain OS I make sure it has it before I buy it. No different
and get idiocy like flash, where it tries to use a gtk related function to check if gtk needs to be loaded?
and as it cant be fixed at the source, one have to build a workaround. and if you start piling up workarounds, you basically get ms windows (where there is, as an example, a workaround specifically for simcity to allow it to access freed up ram)…
this is not politics, this is engineers and kernel hackers being tired of seeing errors that they cant fix. and they cant fix it because its locked away in some black box of binary code…
stop with the IP nonsense already..
amd and intel can give documentation for their stuff, so can nvidia, they just dont feel like it.
selling the hardware with the current driver situation is equivelant to car dealers selling cars with a key genetically encoded to only work for people with white skin, and it isnt right.
only for “software”, freedom suddenly isnt an important thing.
there ought to be laws DEMANDING the release of documentation for hardware, at the same date its put to sale.
If they just don’t feel like it they should be made feeling like it in the name of freedom of choice.
This is a disturbing trend lately.Freedom is defined by a bunch of people and who doesn’t like their definition of freedom ought to be shot on sight.
Might be the same people that praised Nvidia years back for supplying good drivers for Linux when nobody else did.
As for practicality:
With the current state of xorg I rather exercise my freedom to use a binary nvidia driver.
Dude, you took the words out of my mouth. I have seen more and more on the web about forcing people into the GPL with laws, and it disturbs the hell out of me.
I used to think Atlas Shrugged was a load of crap, I don’t know if it is because I am just getting older and my views are changing, but I keep seeing more and more of it going on in the world.
Yeah…enforcing the license a piece of software was released under? THAT’S COMMIE TALK!
i didnt say they should release any code, merely documentation allowing ME, and others, to write code..
stop trying to distort it into forcing gpl..
i didnt say they should release any code, merely documentation allowing ME, and others, to write code..
stop trying to distort it into that “forcing people to GPL”
Sorry, but that is what called free market. And all the companies want to be in a free market, because that would be the best ever. Now that the tables are being turned and they find out what it really means they start to cry. They starting to realize that free market also means that others can choose for you.
So yes I choose against NVidia by not buying there products and going for supplier who’re willing to comply to my demands. One of them is that I want developers to have access to documentation to write a driver. Please note that I’m not saying I’m demanding a driver from NVidia for example.
And to be honest I’m not unhappy to make GPL my golden standard. It has been for over 12 years now and I never felt sorry. It really saved my ass over the years and people near to me are slowly starting to see what I have been talking and doing about. Freedom is more than only code and its something worth fighting for, but in the end you’ll benefit.
Like I’m now having hardware with almost endless software support, or not being monitored will I download new versions of the software I use, or not being forced into buying a new version of Office/Operating System. But also not losing my data since it has been placed behind walls of closed standards, dead code, IP lawsuits.
If that is the price I have to pay to use a driver that gives me 60 FPS instead of 65 FPS then it will be. I’m not going for any short term pleasure when I can get long term happiness. This is one to think about and I advice you to think about it. Making the choice is difficult when you start, but it is rewarding I can assure you. It will also change your life and thinking more then you may expect.
Just my 2 cents
Very well said!
The free market is letting market forces decide where the market goes. What the origional poster was talking about is making it illegal to not choose GPL.
I have no problem with linux (been using it on and off for almost 8 years now), or open source (i frequently contribute patches to the open source libraries I use). I think both are fantastic, and it is great to have an alternative eco-system in the software world. I don’t even have a problem with legislation around open standards for file formats in governments for publicly owned documents and information.
I have HUGE problems with legislation around making the GPL mandatory, as I believe that it is completely inappropriate for large segments of the industry. Let open source compete on its own merits, those merits are there and they are valuable.
In a free market you vote with your dollar. Send NVidia an email letting them know they lost you as a customer because of how they chose not to release a free version of their driver. If enough people do that, they will open up their source. That is a free market at work.
“amd and intel can give documentation for their stuff, so can nvidia, they just dont feel like it. ”
AMD and Intel are some of the most OSS friendly companies out there, AMD is just getting their act together, true, but they are releasing stuff, and Intel has a history of OSS support. Some of the code they use cannot be open by them because they don’t own the IP, but things are much better than they used to be.
Sometimes that is the biggest hurdle for a hardware company to overcome, licensed IP from another company in their drivers. If they don’t own the code, they can’t open it up.
It’s nothing like that at all. Stop the extreme hyperbole. It’s more akin to someone selling cars that cannot run on bio-diesel. If running your car on bio-diesel is important to you, buy a car that runs on bio-diesel, there are several to chose from.
We are at a point where for virtually any class of hardware there is at least one and often several products available with very good Linux support. If you need Linux support buy hardware with Linux support. It really isn’t too hard in the vast majority of cases.
Or you could refuse to buy hardware that doesn’t come with sufficient documentation for your needs, and leave government lawmaking out of it.
History should have made it clear by now that government intervention has, at best, a variable track record when it comes to forcing companies to do things they don’t want to do.
The vendor can focus there budget on developing better hardware instead of the extra expense of developing software.
The vendor can provide a generic driver interface keeping there patent BS within the chips behind the interface instead of in the driver.
Drivers are at a very low level of any OS and bugs or vulnerabilities within them can be addressed much faster under a FOSS license. (thousands of self motivated developers versus a hundred or less “just getting my paycheck” developers who also have all the other needs of the company too support)
Drivers can apear much faster. Moving FOSS drivers from 32bit too 64bit meant recompiling them rather than waiting for some company to eventually get around to recoding completely.
Drivers evolve to a higher level of quality since they don’t depend on budgets that are cut off by the new product line release. The community ATI X module tends to be much faster than the ATI provided X module; with my old 9600 GPU anyhow. Heck, AMD are not providing updated 9600 GPU drivers anymore are they?
If any bit of software code should be under a FOSS license; Drivers most definately are it.
Who’s going to develop the software if not the guy making the hardware? It’s not like you’re going to magically get through OSS a guy who’s willing to work on your software, knows the details, nuances, and bugs of your hardware, and is going to volunteer their time to do the job for free. Even in OSS, people have to write the code and for specialist hardware it’s probably the company that makes the hardware.
Graphics drivers are as big in binary code size as entire OS kernels. Presumably they have a comparable number of lines of code. That’s a pretty big investment to just give out for any and all to see and use when it’s doubtful that anyone will be able to give back due to the skill and knowledge required.
Right, no magic involved, you still need to use real world communication technology like e-mail.
Actually this is exactly what has been proposed:
http://thread.gmane.org/gmane.linux.kernel/487536
The magical part is the existence of a guy or gal who has the depth of knowledge to contribute anything useful to something as optimized as a graphics driver. These people don’t grow on trees and they’re just about as rare and skilled as the hardware engineers building the GPUs themselves. There’s also likely to be a steep learning curve for the field and each particular piece of hardware as well. There’s the magic: finding someone capable of doing this who is not already working at your GPU company or one of your close competitors. Since so much of the differentiation between different GPUs comes in at the softwae level or through clever interfaces, I doubt graphics vendors will want to pool all of their driver resources into one OSS pool and help each other out developing the libraries.
The talent to do this stuff is not widespread so companies obviously want to hoard it for themselves, and it’s not unfair of them to do so (these guys are not behemoths, so we’re only talking about a couple hundred people designing all aspects of their new products).
Do you disagree that this is the reality?
I posted an option where the vendor starts the driver coding then works with community developers; seems that would work. if no one wants too develop support, is the hardware that good? (darwinism would work faster in hardware than it is in software.) also, should’t the profit driver be the hardware rather than enabling interface code between it and OS? it is a general computing component not a specialty device or specialty customer. today, pandering too a single OS is not acceptable unless you make the hardware and embeded OS.
See above. What makes you think that there will be a community of people who can contribute more than typo fixes to the drivers without breaking them?
Yes, and ATI said it again, and again, and agian, and again, that opening up the specifications would be ‘impossible’ – and yet, here we are, with open specifications. It has nothing to do with ‘impossible’, it has to do with a small clique of management and engineers wanting to protect their fyfedom from the ‘hoards of unwashed masses’ getting the specifications and one upping them when it comes to driver quality.
I’ve heard the same excuses over and over, and I wish people would learn the difference between patents and copyright. It is perfectly plausible for something to be open source and patented, just as it is perfectly plausible for it to be an open standard and patented. Developers don’t want source code, they want the specifications of the hardware to allow them to write drivers. The fact that both ATI and Intel have done it, proves that it is not ‘impossible’. It is there fore and issue of Nvidia not wanting to do it rather than not being able to. They’re choosing not to for a political reason, rather than it being a matter that they can’t do it.
And if something is a trade secret? The bottom line is that the hardware and software belong to Nvidia. It is up to them what they do with it. Not the GPL, not the kernel developers.
If the Linux kernel developers don’t like it, I’m sure that they could put code in the kernel to stop the Nvidia (and others) code from inserting itself into the running kernel. This would do several things:
1. Piss the company off and make them stop supporting Linux at all
2. Piss a lot of people off who really don’t care about all of this shyte…
So, the kernel developers freedom ends up removing *my* freedoms (and Nvidias). Tell me again why I should care about the kernel developers? Half of the bastards are on the take from big corporations anyways, that influence the kernel development procedure. It’s a bit like the pot calling the kettle black…
Dave
They are? Really? They are forcing you to use or do something you don’t like? Nonsense.
No ones freedom is removed, you have the choice of not using Linux if you don’t like it and Nvidia arent forced to write drivers for it if they don’t want to.
If they do they have to play by the same rules as everyone else, even if those rules would be “no closed source drivers”.
Hang on here – you and the kernel devs are saying that Nvidia isn’t allowed to write closed binary modules to load in the kernel. I think that is *impinging* on my rights and Nvidia’s rights. Ever heard of a thing called LGPL? That means that you *don’t* have to GPL something just cos you’re linking to GPL’d software.
End of story.
Go have your open source sob story elsewhere, I’m not buying it. I refuse to have my right to run 3D accelerated Nvidia cards on Linux just cos a few fanboy kernel developers want to flex their so called muscles and have a dummy spit.
See, I too can be an a$$hole.
Dave
No, they can write them all they want as long as the Linux license does not forbid it. You do know the difference between urging and forbidding, right?
Why do you think you have that right?
And why do you think I don’t have that right? You’re implying that cos it’s GPL’d it cannot have any interaction with proprietary software (I know that is RMS’s ultimate goal). Well, sadly, as much as I like RMS and the FSF, he’s wrong. Linus’ more pragmatic approach (that the 2 can co-exist) is a much better compromise. It’s one of the few things that I agree with Linus on (other than Gnome blows).
This is so much FSF elitism it isn’t funny, and you know what? It doesn’t endear you to the general public. Instead, they think that you’re some kind of envangelist nutter.
Dave
This probably won’t make me very popular, but … .
Personally I hope that at least one of the usual suspects never open-sources its drivers – Nvidia. I’ve used their binary drivers for many years now. They’ve never failed, they deliver great performance, they are regularly and thoroughly updated and Nvidia are also careful to support legacy cards going to way back when. I just don’t believe the “community” would be able to match this record. More likely, looking at other projects, there would be drivers in a state of permanent betaware, irregularly updated, poorly supported and very likely forked and called IceWeevil when a Debian developer decides he doesn’t like the Nvidia logo.
Yes, open source is a fine thing but not at any price. Linux folks can make a better Linux by improving what they do have rather than by calling for rights over what they don’t have. But I suppose that calling for substantial support for, say, the x-server project, which really needs it, just isn’t so headline-catching
Well said.
If nVidia can produce complete support for there hardware across platforms while addressing issues in a timely fashion then I’m perfectly happy with that as an end user.
I switched away from ATI GPU for the first time in many, many years of building systems specifically because I couldn’t get support across platforms. My view has been tainted a little after trying every howto plus a few archaine rituals in trying to get the 9600 GPU and TV Tuner chip working. The AWI 9600 is just not a cheap enough board to justify that kind of grief. It didn’t help that the binary blob from ATI wouldn’t go above 20 frames per second while the community modules provided full fps; and it was reverse engineered without ATI help.
I still think drivers are one bit of software that should be released with source (it doesn’t have to be GPL, just source available with vendor support). If nVidia can provide X.org with a fully functional module that doesn’t loose development support with the new GPU product line then I’ll use that happily though.
Whilst, yes, nvidia support legacy hardware, it not without problems. Firstly legacy hardware drivers do not get new features even if they possible/work. Fixes are limited to getting it to work with new kernels and xservers and bugs that exist are not fixed.
Binary drivers are never going to integrate well with linux, so-> no kernel mode setting, crappy power management, dodgy suspend/hibernate.
Worst of all is the bad desktop performance. the Ati gpl driver and the intel drivers are very good at xrender acceleration but nvidia is really bad mostly on “new” hardware(only if you count nov. 2006 new).
Now we can all agree that the 3d performance is quite good (the core driver is the same as the windows), but if 2d performance sucks then even with a opengl compositor its gonna be slow.
So with open drivers one will never get to play games on optimized drivers that fully take advantage of the raw computational power of the cards but we would get drivers that are optimised for features linux DESKTOP users care about.
Nvidia can keep providing their binary driver to those that need full 3d power but will then no longer have to support feature for the linux desktop like randr etc.
The gpl driver is improving daily so by the time I get a new computer I’ll buy ati use the gpl driver for linux and for games I can always use windows.
This part of your comment is nothing but border line trolling, if not, a plain braindead snarky retort. It angered me because I know from the later parts of your comment, with regard to Xorg, that you are reasonably well read with regard to the state of the open source Free Desktop and must know what you wrote is stupid.
Iceweasel was not born out of ‘some debian developers’ distaste for Mozilla branding but rather the distribution of Mozilla’s trademarked branding hindering efforts to exercise right already granted under the GPLv2 to every Firefox user to freely modify and/or redistribute the software.
If you want to use ‘Firefox’ on Debian, noone is stopping you. Go download the official binaries or build the tree from source with the branding intact. You can even apply those debian patches. Just don’t distribute it as ‘Firefox’. Seems fair to me.
As for the rest of your comment, yeah NVIDIA do a fantastic job in providing Linux drivers. But there are also downsides to their closed and relatively small Linux development effort.
What about support for new X features? EXA? RandR 1.2? Why are they not working with Intel to bring better video acceleration API’s to the Free Desktop? Kernel-mode modesetting? How about the terrible 2D rendering performance? Go ask a Cairo developer about this last one, or browse NVNews).
None of the above is about slamming on NVIDIA. I love that they provide the high quality drivers they do. But all of the above could theoretically be implemented a lot faster if they open sourced their drivers under the GPL.
Of course there are pros and cons to all development models…
Indeed. But I’ve noticed that when anyone on here points out a few cons, they usually attract posts that are, by your own admission, “angry”.
I’m not in favour of Nvidia going open source simply to people-please, and I certainly think there’s a fair case to make about the extent to which the “community” can match up to the best practices and track record of some commercial software. I can say this about Nvidia because it’s part of the best. I wouldn’t for a moment say it about a lot of other commercial outfits. Trying to paint everything in black and white won’t help take Linux further. There are many grey areas where commercial outfits probably do a better job and I think this is one of them.
As for Debian, I am typing this on Debian and I know the story well. That doesn’t mean I have to agree with everything Debian does. I think that their move to Iceweasel was ill-considered, driven by zealotry and a slap in the face to user-friendliness. Looking at how other distros have handled the issue, I’d say I’m not the only one who thinks so. But that’s just my view, no more or less. And none of it makes me “angry”.
As I was about to answer with an angry post, let me explain why I think your original comment deserves angry answers.
First off, open sourcing something does not imply that nvidia will stop being involved in developing the driver or providing it. It would just mean that it’s open source.
Second, what developers have been asking nvidia and others to do is open up the specs, so that it’s at least possible to write open source drivers for their hardware with a reasonable amount of work. This would clearly not stop nor hinder nvidia from providing their own drivers, be they closed or open sourced, and you’d still be free to use them if you prefer them.
Third, this is not about painting everything black and white, it’s not an ideological issue, though you try to frame it as one, but simply a practical and technical one. Closed source drivers and issues caused by them are simply unmaintainable for kernel developers. This is a very real problem and simply using the word zealot won’t change this.
So,to sum it up, you mischaracterize the nature of the problem and the whole foundation of your argument rests on false premises. I think those are good reasons to get angry about.
I got a better idea, why not linux developers make up their minds an offer a stable platform for drivers where doesn’t matter if the driver is open of closed? You know like Windows Vista did?
Blaming closed source driver is a lame excuse.
I think you missed his point. A closed source driver is the most performant and stable video driver in linux, and has been for years, while open source drivers have been a nightmare.
In the case he was talking about, calling for NVidia to open source their driver is not going to help anyone for practical purposes, only ideological ones, just as ATi opening their driver didn’t help.
That fork had NOTHING to do with the logo, and everything to do with Mozilla/Firefox trademark issues – and the fact that if Debian used custom patches to get Firefox built, they couldn’t use the Firefox/Mozilla branding. Please read up on these things before spouting rubbish.
Having followed the Debian Iceweasel affair closely at the time, I have a strong case for saying it had primarily to do with dog in the manger. Imho, some Debian devs wanted to be difficult and stick it to the big man from Mozilla and found a convenient technicality that let them. Users were effectively ignored.
Were Nvidia drivers open-sourced, it’s a fairly sure bet to me that within a year, some Debian devs would have found a reason to fork the project (even if only a little) under another name. After all, they now have a track record of doing so. Fact. Methinks you do protest too much.
F/OSS = good, everything else = bad is not a substitute for actually thinking. Life simply doesn’t happen in black and white. Were this a one-off from Debian, one should give them the benefit of the doubt. But looking at the ups and downs of the project over the past few years, it isn’t an isolated incident. The project has been marred by irruptions that suggest immaturity and the notion that Debian devs are a class apart.
I’d argue that this rather overconfident approach may well have led not only to Iceweasel but to the recent borked SSL affair too.
Unfortunately, I had a different experience. Since nvidia has several versions of their driver you need to download a very old version to run old video cards (like the Geforce 2 Go I had). The problem is these only support very old kernels so in essence your card will work for 4 years with modern distributions and then just stop. Additionally, overall non-kernel drivers are much more maintenance / troubleshooting to deal with raising the support cost.
Nvidia used to be the default graphics card I recommended to clients and for internal use, now Intel is. Even with the GLX problems Intel has they cost much less to maintain (time and money) and provide an overall good experience. Nvidia is now only used for special high-end needs.
Whereas I do subscribe to the philosophy of open source kernel modules it is the practical reasons (reduced cost) that drive it home for my clients.
You hit the nail right on the head.
I have not had any issues with the Nvidia drivers, providing of course that the underlying headers/kernel src are available etc. Works, works well and is easy to install.
I’m *positive* that this would be ballsed up by the Linux crowd if left to the devices of open source. Personally, I *hope* Nvidia says FU to Linux and stops producing binary drivers. Those that are whinging can stick with the NV driver and get screwed.
The Linux kernel developers have absolutely NO right to demand Nvidia release its driver code. It is proprietary and belongs to Nvidia and is up to Nvidia to do with it as it wants. By taking this stance, they [kernel developers] will only harm the uptake of 3rd party driver developers from producing binary drivers for their hardware. In the end, end users will lose out, because community written drivers either will not exist, or will be badly written and only partly functional. Let’s take Linux back to the stone age…
If the Linux kernel developers are so concerned with their users, maybe they should concentrate more on writing the kernel for end users, rather than huge corporations? The current kernel no longer belongs to the people, but is the slave of big funding corporations, to the detriment of normal everyday users. I personally wish we went back to the earlier days of Linux – things were far better back then imho.
Dave
Meanwhile, these days I only use open source drivers because of Nvidia. They broke my card horribly for several months, and my computer was relatively unusable. I eventually gave up on waiting for them to fix it (4 to 6 months after the problem showed up, no less) and went and bought an older Radeon card, which had open source drivers. It Just Worked for me.
This problem was well known, and there were long threads about it on their forums, but nothing happened about it for at least 2 months after I got rid of their card. Of course, it was an older card, so Nvidia didn’t particularly care about fixing it.
If the driver was open source, I could have most likely worked around the bug somehow, probably with some mails to a development list for help. However, I didn’t have the source, so I was stuck.
Yeah…not at all like the open source, community developed intel driver which is regularly updated, reliable, delivers great performance and supports every Intel adapter available.
Oh, wait…
I’m glad Linux is finally making a push to this route.. OpenBSD (my OS of choice) has been this way forever. Unfortunately, the only issue is the overall weight OpenBSD has is much less than that of Linux…
However, the only issue that still stands is it’s a “suggestion” for manufacturers and not a “NO we are not adding your binary blob in our kernel” – which is considerably different.
All for the better anyways!
First and foremost, is it any surprise the maintainers of the Linux kernel would prefer open source drivers? It’s no surprise that they would issue some kind of position statement on this. Good on them.
This is no way impedes on your freedom to use binary blob drivers. As long as they don’t take steps to hinder binary blobs, they can have whatever positions they want. I’m happy with it.
I know of a least 3 individuals who came down with Mono after playing with the ATI fglrx driver. And there have been reports of several small children who have choked on Broadcom “blobs”. Dangerous? You bet!
Everyone’s argument in favor of OSS drivers is that the community can do better. Really? Considering the lack of support to the X project, could the community really take over driver support for Nvidia and keep up with the 6 month release cycle? Somehow I doubt it. Where are we going to get all these new developers with intense graphics driver experience from? Ohh, thats right. We want Nvidia to still keep doing the work, but open source it. Right. I am a big supporter of OSS, but I am also a fair person. And the group doing the work should get to decide whether or not to open source the software. As a side note, when I installed Fedora 8 on my dell M1330 laptop at the beginning of the year, suspend and hibernate didnt work at all UNTIL I installed the binary Nvidia driver. The OSS nv driver was broken wrt suspend/hibernate. So, lets really think about what we are wishing for before we get it.
How about the people doing the nouvea driver?
And you seem to be missing the point, if we have the source and nvidia drops support for a card or moves support to legacy e.g. no new features even if the hardware could support it, n theory another one could take up the job for maintaining the driver. Without the source code, no one could take up that job.
You claim to be a fair person, okay maybe you can answer me this:
How is it fair that a customer have to change his working hardware, because the vendor has dropped driver support?
This is not just nvidia, they do a good job on their driver by keeping hardware support for old hardware alive in legacy.
I didn’t find it fair when I discovered that my soundcard (SB Live Digtal or something) couldn’t be installered on WinXP after SP2. The hardware worked fine, creative’s driver policy sucked. If I wasn’t a major geek I would have had to buy a new card.
“How is it fair that a customer have to change his working hardware, because the vendor has dropped driver support? “
The Kernel Devs do this as well. I can no longer run my computer using RAID as they removed support for my controller claiming it was deprecated. Funny thing is it is an intel raid controller built on the motherboard, fairly common in desktop systems, and worked fine. Now, thanks to the Kernel Devs deciding what hardware people are allowed to have and removed support, I can no longer run in RAID mode. The computer is only 2 years old. I am not yet replacing it.
The concept of proprietary (closed source) code is diametrically opposed to the concept of Opensource. I can fully see why the kernel developers do not want to include closed source drivers in the kernel. We are not talking about users downloading and installing the Nvidia or ATI drivers, the article is talking about including closed source drivers in the kernel. This would be silly: the kernel is available in its entirety as source code. Does the kernel community start distributing the kernel as they do now AND include seperate files with the source code that are the compiled binary drivers? Why do the hardware manufacturers not distribute Linux drivers with their devices instead? That way they could keep the driver closed source if they wanted to, and users would install these drivers if they needed them.
Does Windows include drivers for every known hardware device? I think not – discs are included with the hardware that include drivers.
I have no issue with the use of closed source drivers with Linux: I do it with the Nvidia driver on my own system, but this is my choice. I am very much opposed to the inclusion of proprietary/closed source drivers in the kernel and am glad the devs are also. If you cannot understand why the inclusion of binary drivers in the kernel would be a bad thing, you do not fully understand the Opensource community and the reason it exists.
I don’t see how this helps the END USER no matter what side your on.
I personally don’t care, I just want my hardware to work, and work well.
I also don’t want to have to recompile my kernel modules every time the kernel is changed.
I believe that this is the final hurdle that Linux must pass in order to be successful for the average user.
Agreed. I see a bunch of people who disagree with the Linux kernel developers getting modded down, so much for freedom of speech!
Thom and others – can you *please* control people inappropriately modding down comments? The rules are specific – if people cannot abide by the modding rules, then their osnews.com accounts should be suspended or removed. These people are abusing the modding rights by not doing the right thing and I’m personally sick and tired of it. We have the Linux zealots, the Mac zealots, the freeBSD/BSD zealots, and if you dare say anything bad about their beloved things, or disagree with them, they mod you down en masse. It is just not right, it’s an abuse of a priviledge, and that priviledge should be removed from their ability to abuse it. Period.
Dave
You forgot the Windows zealots.
hahaha I did to. Sad thing is, they generally don’t muck up too much. They don’t give a **** about the religion of OSes, they just want it to work.
Mac users are always trying to look cool, so they diss anything other than Mac…
Linux users are trying to be “open” so they diss anything that isn’t open…
BSD users are just trying to bag Linux and anything else that doesn’t meet their wants…
Dave
You must be reading a different site from us. Windows zealots are just as lame and moronic as the other OS zealots.
Perhaps there aren’t that many of them but maybe that’s because there’s nothing to be zealous about? I mean, it’s not like anyone actually likes Windows…
Microsoft’s kernel developers don’t have the source code for all of the video drivers from nVidia, ATI, etc. So, how do THEY cope? Well, they have close relationships with nVidia, etc, and they pour significant resources into tracking issues. Why? There are significant marketing opportunities for the graphics card manufacturers to make this level of investment with Microsoft. I think it’s great to ENCOURAGE the driver manufacturers to open their specs; if they do it, the OSS community can leverage their numbers to write new drivers. But what if the driver manufacturers are too slow to make this happen? I mean, seriously, can you blame them? They have limited resources, they’re under pressure to keep moving technology forward, and their primary platforms are Windows and OS X. Linux is kind of a side thing. Let’s be honest about this.
I’ve long felt that the OSS community really lacks an effective lobbying organization. This grassroots crap only gets you so far. GNU isn’ going to cut it. Neither is FSF. No, what OSS needs is a FUNDED lobbying organization that can make certain relationships work better. People could buy memberships in the organization to show their support (kind of like the NRA or whatever). The org could fund development relationships with companies like nVidia, and accelerate the quality of work. But that will take money and effort.
There seems to be a fundamnetal resistance among many OSS disciples to getting into bed with capitalists. But, in my opinion, that’s what it’s going to take to accelerate the pace of change. Bashing nVidia or ATI on the Internet, releasing open letters, yadda, yadda, yadda isn’t going to make them move any faster. But when Microsoft or Apple calls the tune, they listen. Something to think about.
Actually, although I can’t say for sure since I don’t work for Microsoft, large software companies will often have agreements where they will allow each other to view at least parts of their source code under NDA, making it far easier to develop drivers and so on.
So it’s quite likely that Nvidia developers can view large chunks of Microsoft’s code, and Microsoft can view large chunks of Nvidia’s code, for example.
Maybe that’s why there are so many utter crap vendor drivers.
I actually just posted a topic about that on the unixjunkies forum:
http://unixjunkies.ning.com/forum/topic/show?id=2072664%3ATopic…
HW vendors can release closed source code, I just don’t think it should run in kernel space.
That’s why i’ve been big supporter of modular driverspace for linux…it’s best compromise between open source and closed source and that would ensure that no ones freedom is inflicted by the dispute.
Your proposed modular driver space would have direct access to the hardware?If so, then I don’t think that is acceptable either.
We know that HW vendors implement things in kernel space for performance reason mainly but I think that we may be able to find a compromise with a bus-like architecture for certain components that are fully accessible from user space.
Maybe it’s been said but freedom means the freedom to put in any source you want, closed or opened. Otherwise we might as well throw Linux out and go back to Windows. IMHO
I agree. I dont like GNU communitys new “Free as we see it” approach…if i cant put my closed source “binary blob” to my kernel then the kernel isnt open source because it’s closed from me.
Nobody is stopping your from tainting your kernel. They just won’t bother trying help you if you have a problem because of your opaque binary blob obscuring the problem and wasting their time.
It surprises me that very few have read the documents at https://www.linuxfoundation.org/en/Device_driver_statement, which carry a lot of insight into why closed source binary drivers are a bad thing in general for free operating systems.
I would like to add that closed source drivers that are only available for one architecture (generally x86) obviously won’t do much for cross-architecture support. I have PCI Ati Radeon 9200 graphics adapters (and will try to find PCI adapters with newer chips, possibly even Nvidia) in both my SPARC and PPC systems that I hope to get working some time soon.
Without the free Xorg and kernel drivers and reverse engineering projects such as nouveau, I would have to give up on the notion of modern affordable graphics on pretty much anything but x86, which the closed source Ati and Nvidia drivers are nevertheless forcing me into.
Nobody is stopping your from tainting your kernel. They just won’t bother trying help you if you have a problem because of your opaque binary blob obscuring the problem and wasting their time.
If it’s such a waste of time then I’m sure people will move on to something else. The question now is how much freedom really factors in, does it go all the way or only where they want it to go.
Edited 2008-06-25 16:30 UTC