Let me first say, that I perfectly understand the wish for open specs for open drivers. That supporting closed modules is a nightmare, neither Novell nor the kernel folks want to do. And I therefore have no problem in accepting, if they are thrown out because of _these_ reasons. Just call it like this and tell the users "We don't want it, even if it influences your fine experience with our product".
What I'm doing in the following is disputing the arguments that throwing out can be argued as "forced by the GPL". Andreas Hanke schrieb: > Siegbert Baude schrieb: >> Exactly, it should be the choice of the user, not of the distributor. > You are wrong: The user's choice is restricted to what the copyright > holder permits. > > It would be a nightmare if non-contributing users could make decisions > that are against the expressed intention of the copyright holders. This > would render the whole concept of copyright law and also copyleft pointless. As Marcel already pointed out, the expressed intention of the copyright holders is the GPL of the kernel source. Anything else would be just insane with regard to the heterogeneous crowd called kernel developers. Any words on any mailing lists or blogs just don't count. So what was never judged in front of a court was the "derived work" part of the GPL. The interpretation of some kernel folks (again there are also contradicting opinions over there) is (in nice words) very "embracing". They say if the proprietary module uses some header files to attach to the kernel interfaces, this means "derived work". But what, if a vendor produces only one piece for many OSs? Let's say a unified driver for Windows, FreeBSD and Linux? And just adds glue to every OS? Do you think you can really say that this work is derived from the Linux kernel then? And yes there is the "FreeBSD Nvidia driver project" at http://fbsd-nvdriver.sourceforge.net/. Read there: [start quote] Q: Will this alone give me accelerated 3d? A: No, this is the stub component to interface with NVIDIA Corporations kernel, in addition to an NVIDIA card that supported 3d acceleration (TNT or later). [end quote] Wrapping Windows drivers for Linux use is common in WLAN area, too, so there are examples for OS independent proprietary drivers. In my conclusion this says, that the argumentation "proprietary drivers in Linux->derived work" is just flawed. Next argument: Where in the GPL are processor "Ring 0" or something similar mentioned? Where "kernel mode" vs. "user land"? Do you think a judge would follow this differentiation, when kernel folks say "proprietary is fine in user land, but not in kernel mode"? In both cases there is just some interface between hardware and driver, syscalls in one case, direct access in the other. Again, I perfectly understand the technical difference and why the kernel folks very much prefer the one solution, but I'm talking about legal issues now. So if there is no mentioning in the GPL of "kernel mode" or "ring 0" and, as mentioned above, using headers doesn't mean derived work, how could a judge differ between accepted user land and "illegal" kernel mode? He just sees code (hardware drivers) running on a system with the Linux kernel as a glue between user and hardware. But somebody tries to convince him that one case is legal whereas the other is illegal, despite no mentioning of the differences in the relevant license. The only difference is, that the licensor in one case says "it is o.k. for me" (he says so, because it does not disturb his technical integrity of the kernel) that the drivers use the kernel interface in user land, but in the other case he says "it is not o.k. for me" (because the driver can tear the complete kernel down). >From the point of view of the judge however, the difference is just the additional demand from the licensor (the kernel folks). But will he find this in the license under which the work is published? Does the judge know anything about linking code, shared libs, statically compiled in or dynamically used? Should he know or should the license just be enough to decide? Let's furthermore assume somebody just takes the kernel, forks it and says, my interpretation of the GPL is, using header files is no derived work. So you have two identical projects with two identical licenses, but one should be illegal with proprietary drivers and one is not? Do you believe this is sensible? As long as no court has said was is fact in this case and what is not, nobody should argue that the "GPL forces proprietary drivers to be thrown out". And as long as there is no illegal action, also distributing cannot be forbidden. My personal opinion is, there is just an abuse of the GPL in order to force hardware vendors to open specs by social pressure. The GPL never wanted to forbid any use of software together with GPLed code or you could never use any other non-GPL programs on top of your GPL-kernel. The purpose of the GPL is to get modified code back, which is the real interpretation of "derived work". And therefore, if there is some non-derived work it should be o.k. to use it together with GPLed code independent of using syscalls vs. header files. Non-derived in a sensible manner means that the biggest part of the work was done without using anything of GPLed code, which for me is clearly the case for graphics card drivers. >> But to get this choice, the distributor has to deliver the binaries, >> too. > > Novell is much more than just a "distributor". Novell is part of the > kernel community and cannot do things that the rest of the actively > contributing community dislikes without being excluded from the > community in the long term. So you say the real reason is social pressure (or should we call it extortion: "If you don't behave as we like, we make your company life a hell")? > You are totally over-estimating your position as a probably > non-contributing user. (Correct me if this assumption is wrong.) Why should this make any difference for you? And yes, besides bug reporting I'm not a kernel contributor. Does this influence the value of my words? > As such, you have no rights to make any choices other than the ones granted > by the copyright holders. Exactly, and the GPL doesn't forbid distribution of non-derived works, as in my eyes proprietary graphics card drivers are. If one header file is worth more than millions of sophisticated code lines, there is maybe something wrong in the interpretation of the value of code work? At least in my opinion it is. Dispute my arguments from above please, but not with technical wishes based on support reasons or somebody's expressed opinions besides the license iteself, but from a legal point of view, because that is all Novell should care about. Ciao Siegbert --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]