Announcement

Collapse
No announcement yet.

AMD's Hiring Another Open-Source Driver Developer

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

  • #51
    One potential issue wth your specific example is what if AMD were to decide the pot was juicy enough to go for themselves? Do we change the evaluator in that situation? If so, how?

    Comment


    • #52
      Given the real cost of developing the driver in question (think about 10x-20x your estimate) it's a pretty safe bet that we would not be going after the pot.

      That said, it would be easy to add a rule saying that in the event of a conflict of interest (eg same company acting as developer and evaluator) you reserve the right to reject the developer or change the evaluator).
      Test signature

      Comment


      • #53
        Originally posted by RobbieAB View Post
        One potential issue wth your specific example is what if AMD were to decide the pot was juicy enough to go for themselves? Do we change the evaluator in that situation? If so, how?
        Good point. The evaluator is meant to be a third party, which I think would not be the case were AMD to take on the project themselves, being that John Bridgman is a part of AMD.

        But I lean more towards letting the donators democratically decide whether the goal of independence of the evaluator is sufficiently satisfied, rather than formulating some set of rules/laws to which the selection of an evaluator must adhere.
        If everyone involved in the project agree on the terms, I see no reason why we should enforce any type of specific rules on these parties. I suggest we let them decide what is fair.

        Just off the top of my head, I think that after the 1000 donators have agreed to donate the $10 (and thus have agreed on who the evaluator is), their donation is allocated at some temporary location. When a particular developer offers to take on the project, the donators each cast a vote stating whether they accept this combination of developer/evaluator. If at least, for example, 80% of the donators agree that this combination seems reasonable and that no conflicts of interest seem present, the project is accepted. The money that have been set aside are then either paid to the developer when the evaluator deems the project finished or paid back to the donators if the project is not finished within a specific time-frame (specified by the project before it starts).

        I would like to stress that all of these parameters - time to find a developer/evaluator-combination, percentage of donators that should agree on a developer/evaluator-combination for the project to start, deadline for the project etc. - should be specified by a project, and not be inherent to the website in any way. The author of a project decides its rules.

        All of this can quickly become very complicated to explain, and decide on beforehand. But I feel that we'd be able to find a construct that looks after the interests of all parties.

        I'm just basically setting forth the idea of a direct payment of developers by Linux enthusiasts, instead of going through the usual "buy hardware -> hardware vendor hires OSS programmer -> OSS programmer is paid"-route.

        Originally posted by bridgman View Post
        Given the real cost of developing the driver in question (think about 10x-20x your estimate) it's a pretty safe bet that we would not be going after the pot.
        Bear in mind, I have very little knowledge of GPU driver development, not to mention the cost hereof. The example was simply used to create a general idea of the structure of such a project.

        I don't imagine such large projects being proposed in the beginning. In the beginning, smaller projects seem more suitable for this framework while we see if it even works, and which pitfalls/challenges we need to work on.
        A more realistic project could be, I imagine (bear in mind again my almost non-existing knowledge of the complexity of this), "implement shader-based 'Inverse discrete cosine transform (iDCT)' for Gallium3D radeon driver".

        Comment


        • #54
          If the numbers are just supposed to be examples then it all looks good, with the caveat that most of the "not yet done" tasks are more expensive than the things which *have* been done. When seemingly complex features "just appear" in the open source stack what you are often seeing is the last step at the end of months of part-time effort.

          The only other issue to remember is taxes - I remember seeing another similar effort founder on tax problems but I don't remember what the issue was.
          Test signature

          Comment


          • #55
            hmm, interesting idea, we wouldn't happen to have someone knowledgeable enough on the subject to expound on any possible tax implications of that plan would we?

            from my limited knowledge, if you started it as a non-profit, and managed to get your applications to state and federal government for tax exempt status approved (not sure what that entails) that would exempt the organization from income tax, and sales tax, for all money used for the purpose that the organization was granted tax exempt status for.

            employment taxes (and possibly some others) would still have to be payed.

            if someone could get that going i think it would be pretty amazing, and i would certainly be interested in becoming a "donator" as it were.

            Comment


            • #56
              Originally posted by runeks View Post
              "the implementation, in the open source radeon driver, of OpenGL 4, which performs at minimum 50% the speed of the fglrx-implementation".
              It is technically impossible to have an open source OpenGL 4 driver with the current software patent law. Unless you can buy the patent #6,650,327 (floating-point rasterization and framebuffers) and give it to the open source community, open source drivers will always be stuck with OpenGL 2.1 no matter how many full-time programmers will be paid to work on it.

              Comment


              • #57
                "The implementor must be from a country that does not recognize software patents".

                Comment


                • #58
                  Yeah but distributions would not be able to use it anyway, therefore everybody who wants open source GL 3 or 4 would have to compile a 3D driver from source.

                  Comment


                  • #59
                    not neccessarily. legal issues don't seem to have slowed libdvdcss's use down much....

                    Comment


                    • #60
                      Originally posted by AdrenalineJunky View Post
                      not neccessarily. legal issues don't seem to have slowed libdvdcss's use down much....
                      Yeah. Yeah they have. If it weren't for legal issues, every single distro would have DVD playback out of the box. They don't.

                      Comment

                      Working...
                      X