| Weekly edition | Kernel | Security | Distributions | Search |
| Archives | Calendar | Subscribe | Write for LWN | LWN.net FAQ |
Ubuntu Developer Summit report: X.org improvements, driver controversy, andbling (Linux.com)
Ubuntu Developer Summit report: X.org improvements, driver controversy, andbling (Linux.com)
Posted Nov 16, 2006 10:44 UTC (Thu) by grouch (subscriber, #27289)In reply to: Ubuntu Developer Summit report: X.org improvements, driver controversy, andbling (Linux.com) by arjan
Parent article: Ubuntu Developer Summit report: X.org improvements, driver controversy, and bling (Linux.com)
It's up to the lawyers to decide if the kernel maintainers would need to bless/allow this, but if the answer is "they do" then there is a problem for ubuntu ;)
This would be the same problem that Kororaa ran into? IIUC, shipping the same package that is obtainable from nVidia on the same media as the Linux kernel would not be a violation by the distributor. Shipping precompiled binary kernel modules from that package would be a violation by the distributor. "Distributor" would include anyone downstream, as well, such as Uncle Clueless copying his working system onto friends' and relatives' machines.
If such binary kernel modules are derivative works and distribution is a violation of the GPL, aren't the hardware manufacturers guilty of encouraging infringement?
Then again, maybe I'm Uncle Clueless in this situation.
(Log in to post comments)
Ubuntu Developer Summit report: X.org improvements, driver controversy, andbling (Linux.com)
Posted Nov 16, 2006 16:36 UTC (Thu) by notamisfit (subscriber, #40886) [Link]
NVIDIA's in the clear as long as they're not the ones shipping the kernel modules. As for Ubuntu, they've got a cutesy little dodge in which the blob is only linked to kernel code when the system is booted. It's despicable, but it doesn't seem to be a GPL violation.
Ubuntu Developer Summit report: X.org improvements, driver controversy, andbling (Linux.com)
Posted Nov 16, 2006 22:43 UTC (Thu) by lysse (subscriber, #3190) [Link]
> Shipping precompiled binary kernel modules from that package would be a violation by the distributor.
So presumably, running the compiler as part of the installation process for that package would be OK...? After all, a user isn't going to notice the difference (installation means the same thing whatever) but it means that the user chooses to perform the step that distributors cannot, so the distributor isn't distributing a derived work.
...As a law student, I can't help suspecting that a legal distinction that narrow is a bit vulnerable to pressure. If I were a Linux distributor, I'd be loath to do any more for my distributees than point them at where they could get a copy of the drivers and give them really good installation instructions, and maybe a script to do the downloading (and to tell them that if they think this is a problem - and they really should; it *is* a problem - the appropriate people to complain to are nVIDIA for not making something as critical as a kernel module wholly free).
Ubuntu Developer Summit report: X.org improvements, driver controversy, andbling (Linux.com)
Posted Nov 16, 2006 23:02 UTC (Thu) by ajross (subscriber, #4563) [Link]
...As a law studentAs a working programmer with legal opinions, I guess I'm a little surprised at that analysis. Clearly the only people with the standing to complain about Ubuntu's distribution of the NVIDIA kernel binaries (which is clearly within the scope of the NVIDIA license) are the copyright holders of the kernel code. And they (specifically, Linus) have made it abundantly clear, in no uncertain terms, that they do not think that the linking of non-free code into the kernel is not a GPL violation.
Basically, they punted. They aren't condoning the practice, and it's quite clear that they frown upon it. But they clearly aren't agitating for distributors to halt the practice either. So I guess, from a legal perspective, Ubuntu et. al. are "safe" right now. I have a hard time believing that the kernel folks are really going to get snippy about this. They want it to stop, but they don't want to hurt the distro companies.
As a side note: Ubuntu itself does not actually ship the driver in the default installation. The user must proactively download it from the universe repository. So for what it's worth, there's actually a second layer of "protection" there: the user has to choose to install the driver before the "link the module at boot" trick can even be used. There's no click-through license, though.
Ubuntu Developer Summit report: X.org improvements, driver controversy, andbling (Linux.com)
Posted Nov 17, 2006 3:07 UTC (Fri) by notamisfit (subscriber, #40886) [Link]
Actually, at OLS earlier in the year, Greg KH pointed out in no uncertain terms that binary-only kernel modules are, in his opinion, illegal. A quick glance at Google's lkml archives seems to show that he's not the only one. Most distributions that include the kernel modules use some cutesy trick to ensure they're not actually distributing a GPL-violating kernel module; Kooraa's XGL LiveCD got shut down over this one, then came back with Intel and r200/r300 ATI drivers. (FWIW, Linus once made a comment about some uses of binary-only kernel modules not being infringing eleven years ago, and has been living it down ever since.)
Ubuntu Developer Summit report: X.org improvements, driver controversy, andbling (Linux.com)
Posted Nov 22, 2006 14:35 UTC (Wed) by lysse (subscriber, #3190) [Link]
Linus is, of course, only one of the copyright holders of the kernel code; many of the others have expressed the contrary view, and only one of them needs to start a lawsuit.
Ubuntu Developer Summit report: X.org improvements, driver controversy, andbling (Linux.com)
Posted Nov 29, 2006 1:56 UTC (Wed) by dwmw2 (subscriber, #2063) [Link]
Linus is, of course, only one of the copyright holders of the kernel code; many of the others have expressed the contrary view, and only one of them needs to start a lawsuit.... and more than one of them has already expressed an intention to do so ...