Announcement

Collapse
No announcement yet.

Super! Patented Mesa Floating-Point Is Merged!

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

  • #16
    Code:
    configure: WARNING: Floating-point textures enabled.
    configure: WARNING: Please consult docs/patents.txt with your lawyer before building Mesa.
    LOL.
    Note to my lawyer: I didn't actually build mesa, just ran autogen. I swear!

    Comment


    • #17
      Originally posted by mattst88 View Post
      And on that note, I just pushed a change to the x11 overlay that adds a texture-float USE flag.
      I believe you got the USE flag wrong though

      http://bugs.gentoo.org/show_bug.cgi?id=363747

      Comment


      • #18
        Originally posted by DanL View Post
        Code:
        configure: WARNING: Floating-point textures enabled.
        configure: WARNING: Please consult docs/patents.txt with your lawyer before building Mesa.
        LOL.
        Note to my lawyer: I didn't actually build mesa, just ran autogen. I swear!
        an he only remember you to ask you lawyer _before_ you build it ^^. Autogen != build.

        Comment


        • #19
          Originally posted by allquixotic View Post
          This is great news, but it raises a lot of questions.

          Is this essentially the same situation as ffmpeg is in with patented codecs? If so, third party apt and yum repositories will probably start building mesa with these config options turned on before long.

          If "I" (hypothetically) live in and host a server in a country where software patents are outlawed, is it legal for me to distribute compiled binaries of the program containing patented algorithms, to: (a) people who are *also* in countries where software patents are outlawed; and (b) people who are in countries where software patents are *legal*? Further, what is my risk of getting sued in foreign courts by patent holders who accuse me of distributing it to people in category (b), whether or not I ever have?
          Your risk of being sued is probably close to zero. Seriously, it's not in SGI's interest to sue "users" offering this code in their repositories. Have you seen any third party repos being sued by MPEG-LLA? Typically license holders are looking for big fees. Red Hat, on the other hand, would be a compelling target if these were included.

          Finally, would it be possible for someone to go "the Fluendo route" (alluding to their work with their Gstreamer codecs), and legally license the patents in such a way that SGI/S3/Apple/Nvidia/Matrox get their fscking money, while still allowing people to use the latest and greatest Mesa with all the patented algorithms enabled? etc. ...
          IIRC, the Fleundo purchases only licenses the user of the decoder and encoder. Your licenses doesn't extend to the distribution. I.e. if you distribute the licensed codec, you are not able to grant your patent license to the people you distribute the code to. That is why patent licenses are not compatible with free software. Software that is copyleft but patent licensed places addition restrictions on those who receive copies of the software, because they have no license from the patent holder.

          So, if you were to license the code from SGI, build it, and provide gratis access to the public, the public would still not be legally entitled to use the code because the patent license is valid for you only.

          Of course, it's possible for patent holders to grant licenses favourable to free software. For example, I believe the Java patents are granted unilaterally to GPL implementations derived directly from the official source

          Unless they just want to say "no" in order to stonewall free software, and no other reason. I could definitely see Nvidia doing that, and maybe SGI. Apple actually releases and relies upon significant open source software, so I think they might agree to let go of their patent under these terms.
          Considering Apple's position on h264 and their many software patents targeting the mobile platform, I think it's highly unlikely that they (or any other company without a stated commitment to free software) will throw away competitive advantage and give a favourable license to free software. In fact, why would they, when an open source desktop is one of their competitors?

          My last question: Does anyone seriously think that a patent holder on a ClosedGL patent is going to go after end-users for compiling this software themselves, for their own personal use on their own computer, without distributing it to anyone? The number of end-users they would be able to sue would probably be measured in the hundreds or low thousands -- that's not nearly as lucrative as the music and video copyright cartels and the hundreds of millions of users out there who've ever downloaded a song. Considering that a lawsuit in this area costs a ton of money, the lawyers would end up keeping most of it, and the company would get very little out of your average person. With so few people willing to compile Mesa themselves, this doesn't seem likely to ever become a real issue for personal users.
          No, no one actually believes patent holders will be going after individual users. What they do believe is that patent holders will be targeting commercial entities that have a potential revenue stream. This means that those larger companies (like say, Red Hat or Intel) won't be supporting the patented technology and thus won't be spending much developer time on it.

          Without "commercial" support, those technologies will not be part of the default ecosystem, and third-party applications will not be able to rely on it presence. For something like graphics drivers performance, this is a pretty big deal. If open source drivers can only get better performance using unsupported code and (in many places) illegal code, the open source driver becomes much less of a priority for programs that require better performance.

          It's the same with any software patent. You are at a competitive disadvantage if your product requires patented code that you cannot obtain a license for.

          Comment


          • #20
            Originally posted by yesterday View Post
            Considering Apple's position on h264 and their many software patents targeting the mobile platform,
            Apple has very few patents regarding h264. Nokia even has more patents against it as does educational institutions like Columbia. Sometimes I really think people point out apple because "all the other kool people do it".

            Comment


            • #21
              It is really bad when you can not provide mesa binaries with all features enabled. Of course a fast system can compile full mesa in a couple of minutes but do that on a slow one. Also when you want to show something off a live system with a game that requires float it is more or less impossible (well i know only heaven, but that could change). s3tc can be added within a few seconds - the lib is very small, but for float that does not help. The s3tc lib is basically a joke too - most linux gamers need it and because of a stupid patent it is not default. s3 could really officially allow the binary distribution of it, same for sgi and the float extensions.

              Comment


              • #22
                Originally posted by deanjo View Post
                Apple has very few patents regarding h264. Nokia even has more patents against it as does educational institutions like Columbia. Sometimes I really think people point out apple because "all the other kool people do it".
                People "point out" Apple because Apple has specifically involved itself in the discussions regarding the HTML 5 video tag, its possible request of Theora, and the patent problem of H.264. I couldn't comment on Nokia's position because they didn't say anything about this, since they didn't involve themselves/weren't involved in those discussions. I'd expect most people to do the same, that is comment on things they know about, and refrain from talking about what they don't have a clue about. Obviously you have a different opinion.

                Comment


                • #23
                  I did a quick search on floating point textures but couldnt find a succinct (or even a lengthy stand-alone) definition. Could someone either explain it or point to a good source?
                  ***hides head in shame***

                  Comment


                  • #24
                    Originally posted by liam View Post
                    I did a quick search on floating point textures but couldnt find a succinct (or even a lengthy stand-alone) definition. Could someone either explain it or point to a good source?
                    ***hides head in shame***
                    second that

                    Comment


                    • #25
                      Originally posted by liam View Post
                      I did a quick search on floating point textures but couldnt find a succinct (or even a lengthy stand-alone) definition. Could someone either explain it or point to a good source?
                      ***hides head in shame***
                      This is the GL extension added: http://www.opengl.org/registry/specs...ture_float.txt

                      Essentially a texture is just a big map of pixel locations and the associated color value at each one. Pixel[0,0] = 256, Pixel[0,1] = 189, etc. Floating point textures merely allow the values stored inside the texture to be floating point numbers instead of restricting them to whole numbers only.

                      If this concept seems obvious and you're wondering how anyone could get a patent on this idea, you aren't alone. Doing FP calculations in hardware is much more complex than whole numbers, but there's really no difference at all on the software side.

                      Comment


                      • #26
                        I wonder if you need to have the patent to be able to legally use the extension for testing purposes. I'm curious about how much of a performance bump it brings.

                        Comment


                        • #27
                          Float textures may increase visual quality in exchange for performance, but only in apps that can use it!

                          S3TC increases performance by reducing bandwidth load.

                          Comment


                          • #28
                            I haven't read the linked to document yet but it SOUNDS obvious...but that would hardly be the first time an obvious thing has been patented

                            Thanks for the explanation!

                            Comment


                            • #29
                              Large companies that ship Linux with a computer or ship commercial Linux should get a license that will cost the consumer twice of the added license price and with that particular profit buy the patent from SGI or SGI itself.

                              Companies like that should simply be eliminated. The acquired patents should be used to bankrupt/troll other troll companies.

                              Comment


                              • #30
                                Originally posted by marek View Post
                                Float textures may increase visual quality in exchange for performance, but only in apps that can use it!

                                S3TC increases performance by reducing bandwidth load.
                                Ohhh, ok. Thanks for the clarification.

                                Comment

                                Working...
                                X