FSF Issues Fresh Statement Over ZFS On Linux With GPL Enforcement

[gbcox]

One interesting fact which was brought up by RMS was that if indeed Canonical is violating the GPL by including ZFS, they could lose their right to distribute Linux... and because Linux is still under the GPLv2, it would be a permanent loss. "With GPL version 2, published in 1991, a violator automatically, immediately and permanently loses the license to distribute that program. There are no exceptions. To get the rights back, the violator must ask the copyright holders for restoration of rights."

[chithanh]

Well, considering that the CDDL was crafted by SUN's lawyers specifically to prevent code licensed under it to be use on Linux, they must have seen some legal options in enforcing that.

If their lawyers were even only half as good in their jobs as their engineers at theirs, they wouldn't have come up with a license that failed its main objective.

This is not how it works. The license incompatibility is wholly due to GPL clauses. CDDL doesn't forbid combining with code under other licenses. Whether or not the goal was to prevent use of ZFS code under Linux is inconsequential to this question.

Worst case: Oracle sues Canonical. Canonical goes bankrupt.

But Oracle as the ZFS rights holder has no legal standing against combining the code. Only Linux kernel rights holders have grounds to sue.

[Holograph]

That's why Canonical needs to keep ZFS off of their install ISOs and make it a DKMS package that is manually installed or a meta package requirement that gets brought in via an apt upgrade.

See, here's the thing: They don't need to do that. GPL over the kernel has nothing to do with including something in a Linux distro. GPL does not apply to the distro as a whole. A module that is not part of the kernel is a module that is not part of the kernel regardless of whether it's included in the distro install or not. You might as well claim that no non-GPL software can be included with a Linux distro whatsoever, even userspace stuff. DKMS also has nothing to do with this - that controls how the module is distributed and installed and loaded, but a separate kernel module is a separate kernel module with or without DKMS.

FSF and RMS are only hurting Linux by fighting these pointless battles. It was mentioned before that Sony chose FreeBSD for this, and that sure as heck did not help Linux whatsoever. One of the major things holding back Linux from the desktop world is these childish battles between RMS and the rest of the world (and another is the hard-to-configure mess that we call Xorg). Linux would (and eventually will) be better off without RMS. He was helpful decades ago but he's long since gone insane.

Lawyers or not, FSF and RMS are wrong here, and Canonical is 100% right. I don't even like Canonical, but they're right this time. FSF needs to have the guts to admit that they were wrong about this thing and apologize to Canonical. Heck, FSF saying that Canonical is breaking the license here could be considered slander/libel/defamation, so how about that?

[Nelson]

See, here's the thing: They don't need to do that. GPL over the kernel has nothing to do with including something in a Linux distro. GPL does not apply to the distro as a whole. A module that is not part of the kernel is a module that is not part of the kernel regardless of whether it's included in the distro install or not. You might as well claim that no non-GPL software can be included with a Linux distro whatsoever, even userspace stuff. DKMS also has nothing to do with this - that controls how the module is distributed and installed and loaded, but a separate kernel module is a separate kernel module with or without DKMS.

FSF and RMS are only hurting Linux by fighting these pointless battles. It was mentioned before that Sony chose FreeBSD for this, and that sure as heck did not help Linux whatsoever. One of the major things holding back Linux from the desktop world is these childish battles between RMS and the rest of the world (and another is the hard-to-configure mess that we call Xorg). Linux would (and eventually will) be better off without RMS. I'm sure he was helpful decades ago but he's long since gone insane.

Lawyers or not, FSF and RMS are wrong here, and Canonical is 100% right. I don't even like Canonical, but they're right this time. FSF needs to have the guts to admit that they were wrong about this thing and apologize to Canonical. Heck, FSF saying that Canonical is breaking the license here could be considered slander/libel/defamation, so how about that?

Well it is to be determined who is right. the creators of the GPL just clarified what they meant when they had it written. That the LGPL also exists and the subtleties of linking are explained makes me think that it's written such that they are correct. Fundamentally, this is a distribution issue. Canonical are distributing a piece of GPLv2 software (the linux kernel) with included modules that don't follow the GPL, which is more or less exactly what the GPL was designed to prevent. Now maybe there are some tricks where by the zfs is compiled on the user's machine when it's needed and you have to do some tricks to get it, maybe it's downloaded separately or something. So long as the zfs module has to actually be linked in to the kernel there are some questions. There might be some legal wiggle room but it's pretty clear that it goes against the intent of the GPL.

The other piece here, Oracle isn't involved in this and I suspect that they won't be unless they give Canonical some sort of special copyright exemption. The law suit would be between a receiver of ubuntu linux and Canonical, perhaps it would be the FSF that would sue. In fact, as it stands, I'd take the FSF's statements as an indication that they may be planning on it. As the end user of the GPL software, they wouldn't be complying with it as the ZFS module wouldn't be GPL. It could be very interesting if Oracle gave them some sort of restricted copyright exception, that would be something totally new and maybe unaccounted for by the GPL.

[Holograph]

Well it is to be determined who is right. the creators of the GPL just clarified what they meant when they had it written. That the LGPL also exists and the subtleties of linking are explained makes me think that it's written such that they are correct. Fundamentally, this is a distribution issue. Canonical are distributing a piece of GPLv2 software (the linux kernel) with included modules that don't follow the GPL,

Nope, it's 100% clear. Kernel modules are neither compiled with nor linked into the kernel (or is Canonical patching the kernel to include everything? If so then they should change their approach, but can still include it in the install disc.). GPL covers source code, not ABIs. It also mentions nothing about distros, and it's impossible for there to be a legal difference between including it with the install disc or only offering it through the online repos.

FSF may have lawyers, but unfortunately lawyers trained in law, not technology. They're clearly clueless and need to STFU immediately. This fight helps nobody and it just makes Linux look bad/worse. Personally I love Linux and would appreciate it if FSF and Stallman would not ruin that.

Somebody in this thread posted, before my previous post, that it's just like GPU drivers. That person was spot on. Of course, Stallman also complains about GPU drivers and everything else that isn't GPL, so the whining about ZFS is far from new.

Also, as others have noted, Oracle/Sun's license doesn't preclude including it in Linux, so they can't sue for anything. If I drove, say, a Lamborghini, and I lived in a town where cars with more than 6 cylinder engines are banned, Lamborghini wouldn't be able to sue me for driving my car into that town. Same deal here. (I managed to go a whole post without a car analogy, but 2 posts was too much to ask)

[duby229]

Nope, it's 100% clear. Kernel modules are neither compiled with nor linked into the kernel. GPL covers source code, not APIs and ABIs. It also mentions nothing about distros, and it's impossible for there to be a legal difference between including it with the install disc or not.

FSF may have lawyers, but unfortunately lawyers trained in law, not technology. They're clearly clueless and need to STFU immediately. This fight helps nobody and it just makes Linux look bad/worse.

Somebody in this thread posted, before my previous post, that it's just like GPU drivers. That person was spot on.

Look, as much as I hate the BSD licenses, it would have clearly been the right choice. The fact of the matter is if they wanted this driver on linux they should have chosen a compatible license but they didn't. Clearly the fault falls squarely on the ZoL devs. The OSS GPU drivers are written for a compatible license and so obviously fall into a totally different category.

[Holograph]

Look, as much as I hate the BSD licenses, it would have clearly been the right choice. The fact of the matter is if they wanted this driver on linux they should have chosen a compatible license but they didn't. Clearly the fault falls squarely on the ZoL devs. The OSS GPU drivers are written for a compatible license and so obviously fall into a totally different category.

I'm not defending any specific license here. I'm only pointing out that GPL doesn't cover ABIs and that the FSF and RMS have absolutely no case here (and the side-note that wrongly claiming a breach of contract is actually illegal on FSF's part). Unless Canonical is trying to integrate ZFS directly into the kernel, all the module will do is use ABIs defined by the kernel, which are not covered by GPL (the kernel source covering them is under GPL, but using the ABI externally isn't).

I'm also saying that, license or not, FSF should carefully consider whether the things it wants to talk about are helpful to Linux, and if not, they should STFU, even if they are right. How are we going to attact more users to Linux if Linux guys are busy fighting other Linux guys for no benefits?

[duby229]

I'm not defending any specific license here. I'm only pointing out that GPL doesn't cover ABIs and that the FSF and RMS have absolutely no case here (and the side-note that wrongly claiming a breach of contract is actually illegal on FSF's part). Unless Canonical is trying to integrate ZFS directly into the kernel, all the module will do is use ABIs defined by the kernel, which are not covered by GPL (the kernel source covering them is under GPL, but using the ABI externally isn't).

I'm also saying that, license or not, FSF should carefully consider whether the things it wants to talk about are helpful to Linux, and if not, they should STFU, even if they are right.

That's what a number of people are trying to tel you, it is distributed by Canonical with the kernel, therefore it is according to the GPL a derived work. If it was handled by package management nobody here would have anything to say.

[Holograph]

That's what a number of people are trying to tel you, it is distributed by Canonical with the kernel, therefore it is according to the GPL a derived work. If it was handled by package management nobody here would have anything to say.

"With" the kernel package or "inside" the compiled kernel? Does it create a separate file for the module which just happens to be packaged in an archive along with the kernel? That would still be fine - no different than including it as a separate package. GPL doesn't cover that.

If otherwise, then yes, I'll eat my words, and I'm not sure why they would do it that way. You say people are trying to tell me that, but it's REALLY hard to believe they would do it that way (though I do dislike Canonical for a reason, so if anyone was going to screw this up, it'd surely be them). I'd call doing that asinine, because they should just keep it as a module, even if it is included on the install disc (which it should be, if you ask me).

However, I will say this, I apologize if I misunderstood. And if that's the case, again, they should just change their method of inclusion and keep it on the install disc. The first post I quoted was saying that it couldn't be included with the install disc at all, which is false, ridiculous, and the reason I signed up for the forums to comment on.

[duby229]

"With" the kernel or "inside" the compiled kernel? Does it create a separate file for the module which just happens to be packaged in an archive along with the kernel? That would still be fine.

If otherwise, then yes, I'll eat my words, and I'm not sure why they would do it that way. You say people are trying to tell me that, but it's REALLY hard to believe they would do it that way. I'd call doing that asinine, because they should just keep it as a module, even if it is included on the install disc (which it should be, if you ask me).

However, I will say this, I apologize if I misunderstood. And if that's the case, again, they should just change their method of inclusion and keep it on the install disc.

I don't have all the technical knowledge to argue this adequately, but I've read the conversation in a number of places. As far as I can tell there are 2 solutions, 1) change the license to something compatible or 2) distribute it through package management. Only the ZoL devs can choose Option 1) and only Canonical can choose Option 2)

EDIT: The GPL is not at all about protecting the developers. It's about protecting the end user. For the end user the linux kernel is everything that links to it. You just have to realize it's not about the coder, it's all about the user.