How about learning to read?, I never said they should open source _their_ drivers, I don't care what they do with their drivers, I said they should release the documentation so I can _use_ the hardware I paid for.

And to say that there is IP that they can release about how their shitty NICs work is too stupid to be even funny, they are the only company for years that has refused to release such documentation.

Anyway, binary drivers have no place in Linux, if I wanted binary-driver-hell I would run Windows.

Funny we don't ever see any ATI community interaction. Kudos to nVidia!

I will send the question regarding releasing hardware specs as a followup right now. Stay tuned.

--jeremy

Im not exactly happy with the "no comment" response to the X.org question.

Nvidia Open Source driver works just as well as any other video driver, with fully functional Xv support. So really, we get the best of both worlds... Believe me, when a open 3D chip with decent performance comes out, I'll be the first in line to buy it, but until then..

I'd like to thank Nvidia for producing FreeBSD drivers too! We have several Dell desktops which we specifically chose because the the graphics chip is supported. I know this is a Linux forum, but thanks anyway.

I totally subscribe what you say, and shocked about other's comments about "the need of let binary closed souce driver be part of [GNU/]Linux distributions".
This way the Freedom will end really soon. If you change master, you are not more free.
The documentation is the key, and having people avoid buy hardware that is not FreeSoftware friendly, and NVidia is not.

Sorry, but markit, could you re-type that, i have no idea what you just tried to type

You say this is the way freedom will end....so what...first they will release binary drivers in an elaborate scheme to take over the Linux kernel? No one said you have to use them.

In regard to Nvidia not publishing the documentation to their chips I have a few comments.

All video chip manufacturers have been tight lipped about hardware documentation since IBM published the SVGA programing specs. Often this information is only available under NDA and often they will only give it to you if they deemed you really did need it and your needs benefit them while not giving any information out. Even getting them to cough up Documents on obsolete chips is impossible ( I know this from working on the Bochs IA-32 Emulator Project).

At work I develop wireless communications equipment. I am seeing this "we won't tell you how to talk to it" approach more and more in all sorts of chips. They are even applying this policy to people trying to build stuff using their chips. At most I can only get a pin out of the chip, the voltages and currents it needs, and maybe some crude documentation on the bus interface. Our company has even had companies refusing to sell them chips because we are too small ( we fill a niche market ).

Most recently, our latest product was designed around the IBM 405EP embedded micro processor. IBM was excellent for providing documentation on this, you could download all the technical details on the chip without having to get an account or sign an NDA. Unfortunately IBM sold off this processor line to AMCC. AMCC is treating all this information differently. They will give out a few bits of information on this part, a broacher, an application note showing how it could be used in a system, and a data sheet with mainly physical specifications of the part. To get a programming manual you must get it under a NDA. We use linux as our operating system on our latest products, and not being able to discuss programming details, problems, bugs and such with other linux people is a _big_ hinderance. Needless to say we will be using different processor made by a different company in all new products from here on out. I can understand and live with be under NDA for advanced (preproduction) information, however once a chip has been announced and is in production it is a different story. Note that AMCC is not the only company that operates like this that we deal with, they are just the most recent one and fortunately we have the opportunity to go somewhere else.

NVidia wants to try and be ahead of the game as far as what their hardware does and one of the ways to do that is to not let competitors get detailed information about it. It's as simple as that, even if you do see some open source 3D chip appear in the market it will most likely be 2nd or 3rd generation technology because giving out your best company secrets is to a hardware company like just handing your books over to the competition. This is why it's so hard for companies to agree on and establish standards like PCI or USB or any other technology that took forever to come about because it involves more than one company.

We've found ways to make open source software work only because there are other means for developers to generate revenue from it but hardware isn't there yet and may not be for a long time, if ever. Having the specs for the direct software interface to your hardware is almost as good as having the hardware specs and companies don't want that, thus the machavelian Non-Disclosure Agreements that Obscene_CNN mentioned. All of the hardware companies are ambiguous to releasing their programming specs, some more than others, these guys have just done one of the better jobs of actually implimenting a highly functional closed-source solution rather than just thumbing their noses at the open source community altogether. The fact that this differs from the philosophy of open source seems almost irrelevant seeing as how they actually work on a solution. Their drivers themselves still need to be compiled against your kernel, so rather than thinking of it as closed source 'taint' think of it as a small shift of what you call 'hardware' to include some code.

Or maybe we should just ask them to give us the hardware for free too?

Darin,

Actually with the rapid movement in the video chip market I don't think releasing detailed programing specs would help a competitor much. By the time a competing video chip maker could reverse engineer the spec, incorporate the idea into their design, and make a chip two more chip generations will have come out.

The complexity of todays video chips is such that a programing spec is not going to give away all the critical details of the chip inards. They are similarly to high end microprocessors. The secret sauce is how you pipeline the execution unit and an efficient memory interface, not what instructions you feed it. If video chip makers come up with a break through in technology they get a patent which locks out the competition. It is true that to write good *fast* drivers you need to know details on the internal pipelining of the chip, however to write functional drivers you do not need to know this. Microprocessor manufactures give out programing details on their chips (except the dorks at AMCC) and we don't see hoards of clones out there. Today there is no good reason to keep programing information secret.

That's not entirely correct. The Simputer(tm) www.simputer.org for example is a GPL'ed palmtop computer that has been openly available for 4 years.

Nephilus- that Simputer looks like a handy device, I imagine there will be exceptions to every rule though.

Obscene_CNN - I agree with you about programming specs vs internal workings except product development involves writing drivers prior to hardware release (At least in the Win world, as it should in the Linux world) so the time gap is not all that big if specs are released before the hardware.

I guess my view is affected by the M$ world where hardware vendors provide the drivers. This gives the advantage that the coders know how the h/w works. If more vendors put effort into drivers that NVidia appears to have put into Linux, even in a closed implimentation, we would see more h/w that works well in Linux and works better than if drivers were written by people less familiar with the h/w.

Open-Source drivers are almost pointless if the vendors really work on effective Linux drivers. In the end the consumers choose the h/w mainly because it works. Imagine a vendor saying they won't do a Win driver and handing M$ the programming specs, would anyone praise them more? It could also be argued that there is internal software in the chips, should that too be Open-Sourced? The Linux community is growing and the typical Linux user is not worried about 'true' Open Source, they want h/w that works and works well. If this involves closed drivers, that I think the vendor should be responsible for anyhow, then I say kudos.

There are several views that hardware manufacturers have taken on drivers for the Linux community and I think the true test will be whether it works, not how it was implimented. If their stance is to put real effort into drivers but don't open the code then let others come forth and do a better job and we can let the market decide. In the video card market has any other manufacturer come close to the level of commitment that NVidia has?

I like the Open Source philosophy and I don't think their method is perfect but I still praise NVidia for doing something that works.

Linux supports _dozens_ of architectures, and there are dozens of Open Source operating systems that are not Linux. Anyone that runs either is locked out from using the hardware they paid for because NVidia wont release the needed documentation to _access_ their hardware.

The reason people runs Open Source operating systems is to not get locked (in or out), and not for religious reasons, but for purely and sound business reasons; using hardware which only has binary drivers is throwing all this advantages out the window.

I run the tech department of a small/medium size company; and I have been screwed over by both software and hardware companies that wouldn't want to tell me how their junk works, just so they could bait and switch on me; force me to "upgrade" to versions I didn't need, pay insane amounts of money for documentation to use the hardware I had paid for, etc.

The argument that that information would help their competition is bullshit, because we are not talking about the inner workings of the chip, only about the _interface_ to access it, and anyway their competition has more than enough resources to reverse engineer their binary drivers drivers, and I'm sure they already do.

Back in the dark ages network card manufacturers didn't release documentation for their NICs, this changed long ago, and this days only NVidia(and other really stupid companies that fortunately aren't very popular) refuses to release docs for their NICs.

It's about time graphic card manufacturers wake up and get out of the dark ages. Releasing the documentation benefits their customers, which directly benefits them, as making your customers happy is a good way to have them come back. And by customer here I'm not talking about an 31137 kid playing DooMIII at his mom basement, I'm talking about companies buying tons of hardware to get real work done.