Announcement

Collapse
No announcement yet.

Gordon's Thoughts On Open-Source GPU Drivers

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #61
    Originally posted by RealNC View Post
    According to whom? You? The Linux kernel devs have made it clear that it's not about being Free, but about being awesome.

    I couldn't care less about the "GNU" in "GNU/Linux". I only care about the "Linux" part. GNU can go hump themselves for all I care.
    Then please rm -f the following (very incomplete list):

    bash, ar, gcc, g++, ar, cpp, ld, nm, as, ranlib, objcopy, readelf, libiberty.{a,so}, libc.so.*, libgcc_s.so.*, libstdc++.so.*.

    Have fun getting a decently-running distro going with clang, newlib and ksh. Most stuff doesn't even compile with clang, and even if it did, glibc-specific behavior is relied upon quite extensively in many programs. Not to mention all the bash-specific system administration scripts that constitute the convenience of popular distros such as Ubuntu, Fedora, OpenSUSE and Gentoo...

    Seriously, incendiary comments like this are completely counterproductive. If we take you literally, you look like an idiot. If you take yourself literally, you have a non-working system with a small subset of all the FOSS programs operating correctly (not to speak of proprietary programs) -- have fun with that. And if you were just making hyperbole, while in fact being fine with using GNU software, you still look like an idiot: if you care this little about what GNU has done (and why GNU has done), you are essentially putting all your faith in the current state of GNU software. If all maintenance on GNU software stopped today, and all developers who contribute to GNU who believe in freedom suddenly stopped contributing, the future of all those programs would be in jeopardy. Bitrot would take its toll, and these programs would end up in a very poor state as other software evolved around them. Is that what you want?

    The development of free software (or open source, or "software I can take from others without paying them anything", or whatever you want to call it) depends crucially upon a positive feedback loop: share, contribute, use, repeat ad infinitum. The only thing preventing the entire ecosystem from becoming proprietarized in a short and tidy corporate coup is the fact that people value freedom and/or the availability of source code, which is at the very crux of both the Free Software and the Open Source movements. People who release software under licenses that help ensure that the source code and essential freedoms are preserved in perpetuity, are doing so because they want the positive feedback loop of progress to continue.

    I don't know what you call the philosophy of "I like software that is awesome and I don't care about freedom or the availability of source code", but that is not the philosophy of any Linux kernel maintainer I know of, not to speak of other FOSS projects. That statement is neither in support of Free Software, nor Open Source. Andrew Morton is pretty stalwartly a Free Software guy; Linus Torvalds is pretty much an Open Source guy. Both of them value the perpetual availability of source code, not "just" awesome software. So don't put into the mouths of the kernel maintainers something that they do not profess, just because that happens to be your viewpoint.

    BTW, think on this for a while: a developer contributing to a FOSS project who does not value the perpetual availability of source code on any level is absurd. If they don't value it, why are they contributing their code in the open? Because the license of the software legally forces them to release their changes? So they are so ethically destitute that they will just blindly abide by the license of the software that someone else decided for them, without considering whether they agree with it or not? I don't buy it.

    And no, don't talk about employees of companies who hate FOSS but contribute to Linux anyway because they get paid. The company, who is the real operating agent here, cares about either freedom or open source -- otherwise they wouldn't be contributing to an open source kernel.

    You could say that the company is self-interested and just wants to make profits today, but this is just a present vs future argument. Of course they want to make money right now, but they know that the open source ecosystem helps them make money into the future as well. If they thought otherwise, they could (and probably would) use a proprietary OS. A major investment such as hiring engineers to work on the Linux kernel requires significant corporate planning and research into future prospects; they wouldn't just make this decision on a whim. And in their estimation, they must have determined somehow that their profits will be higher if they invest in open source than if they go with a proprietary alternative. Otherwise they would have been financially motivated to go with the proprietary alternative.

    So I've taken us from valuing freedom, all the way through companies contributing to open source because it is economically lucrative for them. But regardless of whether you are Richard Stallman, Linus Torvalds, or an Oracle Linux engineer, there is a common theme and assertion being made: we care that everyone's shared source code, our community body of work on "Linux", is available today, tonight, tomorrow, next year, and in 5 years. And we would be very upset if a large portion of new work on Linux, hugely beneficial to everyone, were written under a proprietary license. So we value the perpetual availability of the source : not just right at this instant, but into the future as well.

    But wait, what would prevent someone from doing precisely what I just said? That is, what prevents $FOO_CORP from writing some massive enhancements to the Linux kernel, beneficial to everyone, and not sharing them with anyone but their paying customers (who get a binary blob)?

    Well, unless you're talking specifically about the proprietary video drivers (which are perhaps a real-life instance of this example), the answer is "the GPL". So you can see how everyone who put their trust in the ecosystem, that any major new enhancements to come down the pipe would be shared, has thus been cheated by the proprietary drivers: everyone from the freedom-believer, all the way down to the economist at Fujitsu, has been cheated. Because they all expected that this work would be open source, just like the rest of Linux.

    And how is it that they were cheated? Because, arguably, the GPL has a loophole that permits (or does it?) the proprietary blobs to be written. If indeed such loophole exists and is defensible in court (which has not actually been tried), this is unfortunate for everyone, because all stakeholders involved are not getting what they expected. The license is supposed to protect us from this. It didn't. It's a mistake, an aberration, an exploit, or maybe what they're doing with the binary blobs is actually unlawful -- call it what you like.

    The reason people say "if you don't value freedom then use Windows" is because, well, using GNU or Linux (or GNU + Linux) directly implies that you value one or more of the following:
    (a) freedom in and of itself,
    (b) the availability of source code in and of itself,
    (c) the quality of software products produced by a healthy FOSS ecosystem composed of developers who value either (a) or (b).

    Even if you value (c), you must indirectly value either (a) or (b) by causation: because what caused all this great FOSS to be written? Was it... cosmic rays? An asteroid? A lot of bored people? No -- it was people who very purposefully evaluated either ethical or financial aspects of either free or open source software, and determined that they want to spend their time contributing to this positive feedback loop of progress.

    So if you don't, in fact, value the very engine (or ecosystem, or feedback loop) that enables the creation of this software, then you are basically a nearsighted fool. By "not valuing it", that means you basically wouldn't care if the entire ecosystem dropped off the face of the earth tomorrow. And what would remain would be whatever software you currently have downloaded onto your PC, and nothing more. And you'd be perfectly fine with maintaining and enhancing this software yourself, in complete absence of the ecosystem that originally produced it. And I don't think that anyone is so foolish as to actually hold this viewpoint, except perhaps the few people who understand every last subsystem of Linux and actually could maintain the whole thing alone (albeit at a much slower pace!)

    I'm not exclusively in "the free software camp", nor am I just an "open source advocate". But there are commonalities and pragmatic agreements of fact between both movements. I have attempted to pull out these commonalities and show how they are the very genesis of not just GNU, but all software released under a free or open source license, your precious Linux included. If you don't care what will happen to your software tomorrow or next month, then I dunno -- maybe you're a terminal cancer patient just looking for a few last laughs before you pass on? Because most humans look into the future with concern about what may come to pass. And without the free and open ecosystem, and the principles that underlie it, none of this would be possible.

    Comment


    • #62
      I am ignoring all other posts in this thread, because, frankly, they've kind of gone to shit.

      SGI *did* bring legal action against ATI for these patents; this isn't some idle threat that can be ignored. If Icculus or anybody else feels like they can offer some serious assurance to distros that they and their users will not be the target of similar lawsuits, they can do that on their own time, rather than blame Mesa's maintainers.

      Also, if Icculus feels that he's got something to bring to the table, he can always email the mesa-dev mailing list, which will get our attention much quicker.

      Comment


      • #63
        Originally posted by deanjo View Post
        Sure they do, but why even buy a discreet card if all they are going to do is "basic functionality" when any integrated solution could carry out those duties quite readily.
        This is like saying all my mother ever does on her computer is checking email, so there's no need for a quad-core cpu. Therefore, if she buys a computer with a quad-core, it won't work. Because clearly she doesn't need that hardware, so why did she even buy it? I'll tell you why, it's because she has no idea what she's doing. She saw a marketing sticker telling her how awesome it was, and thought sure, i can afford that. So why not?

        Or how about this scenario - i buy a computer with a decent graphics card and use it with windows (or binary drivers under linux). A few years later, the hardware has gotten old/slow enough that i decide to buy a new computer and use it as my primary machine. I re-purpose the old one into a linux machine that just does basic tasks like email/web/compiz. Why shouldn't i be able to use that? I'd be stuck just because i bought a decent computer a few years ago that didn't include intel graphics?

        I can come up with these scenarios all day long, there are millions of them. The bottom line is that all commonly used hardware should be supported at a minimal baseline level. Those that need more can look into getting it, but you can't punish people for having better hardware. That's just stupid.

        Comment


        • #64
          Originally posted by MostAwesomeDude View Post
          I am ignoring all other posts in this thread, because, frankly, they've kind of gone to shit.
          That's probably smart.

          SGI *did* bring legal action against ATI for these patents; this isn't some idle threat that can be ignored.
          Didn't they lose that lawsuit?

          If Icculus or anybody else feels like they can offer some serious assurance to distros that they and their users will not be the target of similar lawsuits, they can do that on their own time, rather than blame Mesa's maintainers.
          The code is already present on the same server. Hiding it behind a configure flag so that it's never actually compiled is EXACTLY the same as the current situation. I've never seen any legal argument otherwise. And I've never even heard of any other project that had any doubts about this. See Freetype, for example. Or basically any other project that ever had to deal with patents.

          Also, if Icculus feels that he's got something to bring to the table, he can always email the mesa-dev mailing list, which will get our attention much quicker.
          I agree that his post was ridiculous and won't accomplish anything. On the other hand, it's pretty clear by this point that posting on the mesa-dev mailing list won't either. So in the end i don't think it really mattered.

          Comment


          • #65
            Originally posted by deanjo View Post
            Sure they do, but why even buy a discreet card if all they are going to do is "basic functionality" when any integrated solution could carry out those duties quite readily.
            .
            Because lots of people buy hardware without having a clue, lots of people get hardware via jobs or otherwise, lots of people have PCs with no integrated video and just buy a GPU, or get it from Dell or whatever.

            Like really you should probably do a small bit of thinking before posting so the rest of us don't have to point out the stupid.

            it would make these forums a lot easier to read though if everyone did that.

            Dave.

            Comment


            • #66
              I've read a lot of good points here and a lot of...flaming, but here's a thought that maybe some people could get behind. The proprietary drivers are a nice "stopgap" for gamers, professional graphics workstation users, and what not until the open source drivers can one day match or beat their performance. This may take many years, partly because it may require either an abolition of software patents or a new understanding by patent holders that makes them exempt FOSS software for FOSS operating systems, like mesa. The latter solution would probably require some serious force by some powerful firms like Google in a way that ultimately would cause a market shift from the way things are now. I like to think that that day will come one day. Then, we could all look forward to the blobs going away and the open source drivers replacing them completely. because they are as good or better.

              Comment


              • #67
                Originally posted by efikkan View Post
                nVidia's market share has nothing to do with this, I suggest you re-read my post, especially the second paragraph. The claim that only gaming enthusiasts would greatly benefit from nVidia's proprietary drivers is not true at all.
                I'm not making that claim (which, as you say, would be false). I'm rather questioning your claim that "most users" or "most of us" benefit from Nvidia GPU driver features, considering that most users don't use Nvidia GPUs in the first place and that Nvidia doesn't contribute significantly to the shared Linux GPU driver infrastructure. It's all well and good that Nvidia users benefit from Nvidia drivers, but that doesn't mean everyone should kowtow to Nvidia and pretend their drivers are some magical ideal with no drawbacks.

                Comment


                • #68
                  Originally posted by Ryan Gordon View Post
                  Also, I find it completely ridiculous that we're shipping open source OpenGL drivers without S3TC support because of patent concerns. Today, that's like shipping a web browser without .jpg support!
                  Or like using the GIF file format in the 90's and hoping you don't get sued...

                  http://replay.waybackmachine.org/200...9809071211.htm

                  Or like delivering a web browser in the modern era without h.264 support...

                  http://blog.chromium.org/2011/01/htm...in-chrome.html

                  I do wonder if Ryan was referring to the GIF patent issue and got the format wrong, and if so whether he thinks that the GIF issue was solved by somehow safely ignoring the patent, rather than having to actually outlive the patent.

                  Comment


                  • #69
                    Originally posted by deanjo View Post
                    Hey I have no problem with people using open drivers if it meets their needs. What I do however mind is opensource fanatics trying to pull a Obi-wan Kenobi, waving their hand in the air saying "These are not the droids you are looking for" in respects to driver capabilities and saying that they will cover most peoples needs as the open drivers are lacking in many areas to exploit the hardware capabilities that they paid for. If basic functionality is all that is needed then really the opensource driver developers shouldn't even be looking to add discrete card support as an integrated graphics solution should be able to cover what according to opensource fanatics is "most peoples needs".
                    Nobody is preventing you from using your closed blob.

                    This crusade has always been about you trying to prevent people from having a truly open source system because of your deep conviction that nobody needs, wants, or should be allowed to use a fully open source system.

                    If you cared about functionality, you'd be using your blob, or using Windows, or using your Xbox, but here you are on a Linux forum shitting on open drivers like there's not tomorrow.

                    We'll leave your blob to you, you leave the Linux decisions to RedHat, Fedora, SuSE, Debian, and the others who actually care. Stop telling those who develop Linux how they should develop it.

                    Comment


                    • #70
                      Originally posted by efikkan View Post
                      We have to stop criticizing nVidia for their GNU/Linux drivers.
                      We are not criticisint Nvidia because of their drivers, but because they refuse to release specs.

                      Comment


                      • #71
                        Originally posted by efikkan View Post
                        nVidia's market share has nothing to do with this, I suggest you re-read my post, especially the second paragraph. The claim that only gaming enthusiasts would greatly benefit from nVidia's proprietary drivers is not true at all.

                        For instance, a low end card like GT 430 (or maybe even a cheaper card) with proprietary drivers will offer video acceleration, great power management, able to run many games, offer OpenCL support, as well as acceleration for compiz, firefox and much more. With the open source drivers the computer would need a much more powerful card to even come close to the performance, still lacking decent power management so the card will be a lot hotter and drain more power, and lacking video acceleration (wich almost anyone is interested in) so the computer would need a more powerful CPU.
                        Have Nvidia release the specs, and the problem will be mostly solved.

                        Comment


                        • #72
                          Originally posted by locovaca View Post
                          And perhaps the issue is not with the closed source drivers but with a Kernel and Xorg API that changes multiple times a year.
                          It's a Linux not Windows or OSX where known vulnerables aren't fixed for months. Open drivers work with the every new kernel and xorg releases, so I don't believe it's so hard to make blobs working too.

                          Comment


                          • #73
                            Originally posted by locovaca View Post
                            Hey, I'm just calling it like I see it. From a professional development standpoint, it makes Linux look more like a toy and a hobby it started out as and not a real system to target when the goal posts move every 6 months.
                            It rather looks like an approach to better solutions. There's LTS Kubuntu release, so you have a stable system for three years. Try the same with Windows. Its service packs break stuff very often. I am able to play some windows games on Linux using wine while they no longer launch on windows.

                            Comment


                            • #74
                              Originally posted by numasan View Post
                              I also smell a lot of hypocrisy here, with people flaming others for not sharing their purist views, but still cheering for "closed" companies like Unigine and Valve for considering Linux, or hail AMD for opening technical specs while still using the Catalyst driver to play aforementioned proprietary games. And if you deny any closed drivers, why are you even interested in what closed source software developers choose and do? Probably because you want to get the full potential out of that supercomputer of a GPU that AMD sells.
                              Driver is a part of the system while some game is not.

                              Comment


                              • #75
                                Reading all this stuff where some people prise blobs and nvidia I wouldn't be surprised if nvidia fears AMD open source strategy. OSS drivers will give enormous advantage over nvidia blobs when they'll be mature enough. Think of Gallium etc. Amazing stuff.

                                Comment

                                Working...
                                X