Canonical accused of violating GPL with ZFS-in-Ubuntu 16.04 plan The Software Freedom Conservancy (SFC) thinks Canonical, the curator of Ubuntu, has breached the Gnu General Public Licence (GPL). As the Conservancy explains, Canonical recently announced that Ubuntu 16.04 will “make OpenZFS available on every Ubuntu system. Canonical reckons that adding OpenZFS represents “one of the most …
Just use FreeBSD.
Or just download and compile ZFS for yourself now.
My Ubuntu 14.04 fileserver runs just fine with it.
"This appears to be Sun's fault, and now Oracle's fault, because they used a derivative of a well-recognized public license, the MPL, a clear violation of the 'must be literally blessed by a man who appears in public wearing a disk platter on his head under the alter ego of "St. IGNUtius"' doctrine."
Sun devised this license on purpose to be as free as possible while remaining incompatible with GPL and nobody said they didn't have the right to do that. What SFC tries to do here is to prevent a legal confrontation. I hope you agree with me that Canonical is naive to think they can commit this violation and get away with it. Oracle's lawyers will tear them into pieces.
"Canonical is naïve to think they can commit this violation and get away with it. Oracle's lawyers will tear them into pieces".
This.
I don't think it would matter even if they win in the end because Oracle have the resources to bleed them dry in the process.
While it's been a while since I used Ubuntu directly, I do use distros derived from theirs. And, anyway, diversity is a good thing to maintain regardless. So I would very much not like to see them buried under a huge pile of legalese.
I hope you agree with me that Canonical is naive to think they can commit this violation and get away with it. Oracle's lawyers will tear them into pieces.
Nope, you've got the wrong end of the stick entirely. Oracle don't and won't give a damn. They didn't when FreeBSD incorporated ZFS, and they won't here.
No, it's the GPListas and the kernel devs who may get cross, but Canonical's lawyers think that they have no reason to do so. The trouble lies partly in the fact that GPL2 has not really been tested conclusively in a court case in this area.
Anyway the whole thing is nuts, and it's only the foamy mouthed zealots who care. ZFS is a fine bit of code that everyone wants to use, and it is open source.
The Linux crowd's normal response to this sort of problem is to reproduce the software; they did this with DTRACE, creating FTRACE. However they have failed to reproduce ZFS satisfactorily. The ongoing lack of ZFS or a decent reproduction of it in Linux is making Linux look bad.
I think Canonical are being quite brave and are trying to move Linux on for the benefit of all. We should applaud that. Incorporating ZFS will not harm anyone or make any existing or future code more or less open. There's loads of people who are compiling their own ZFS.ko anyway, and obstructing Canonical would be peevishness itself.
Personally I think that the clauses in GPL2 that force GPL2 onto derivative works have become a big obstacle to progress. If they were updated to permit use of other acceptable open source licenses too then there'd be no real problems. The GPL2 is just words, not a sacrosanct document that mere mortals cannot change.
The same goes for any other restrictive document such as the US constitution and magna carta, both of which have been amended and or partly repealed. Good grief, if even US politicians can occasionally agree on amending the constitution, how bad does that make the GPListas look?
Spot on. The problem is that the Linux kernel folks view compiling a kernel module as "creating a derivative work" of the Linux kernel. It is a fairly fine line - the GNU folks have always regarded linking a library into a piece of software as creating a derivative work of that library and I believe have won this point in court, though I can't point to the case off the top of my head. This is why there is also the LGPL, which allows creating this sort of derivative work without forcing the (L)GPL onto the derivative.
The kernel has no such exception for linking, so the question becomes, when does loading a kernel module cross the line into linking a dynamic library? There are arguments both ways here.
when does loading a kernel module cross the line into linking a dynamic library?
I find the logic of the SFC questionable here.
There are many binary modules used on a typical desktop Linux system. Graphics card drivers from the manufacturer are almost always binary and non-free. I would love to know what the difference between a distribution supplying Nvidia and AMD graphics drivers and them supplying the ZFS kernel modules are...
creating a derivative work
This is a bit of a mire. There's some rather elderly discussion here, but I'm pretty sure the linking issue (or at least, the dynamic linking issue) remains moot.The only cases I can quickly find reference to where there has been a determination are the more flagrant ones.
The first so-called test case of the GP fizzled out. Interestingly, Rob Landley, one of the BusyBox developers later claimed that the court cases had made BusyBox too dangerous to use, rather defeating the object of open source software and has created ToyBox under a more permissive licence.
I suspect that the dynamic linking reach of the GPL could have been defeated by the following logic.
1/ Define an abstract API that provides the necessary additional functions for your software.
2/ Distribute your software on condition that the recipient provides a library to match the API.
However, now that APIs have been determined to be copyright (in the Android case), it would be interesting to see the SFLC lining up behind Oracle so close that particular loophole.
Of course, GPLv3 is more proscriptive about what it considers a derivative work. However, the licence doesn't really get to determine what a derivative work is - that's defined by copyright law - so exactly how much of it really applies remains to be seen.
And that's really the problem with the GPL - it may be well intentioned, but it's the same impenetrable and potentially unenforceable guff as you found in any click-through or shrink-wrap licence from a commercial company. It's a bunch of people try to protect what they see as their own interests -and just like "for your own comfort and safety", pretending it's in your interests too.
Of course, people have the right to licence their software as they please, but if it becomes mired in legal uncertainty it's ultimately self-defeating.
I think SFC are wrong here. Linux is GPLv2 rather than GPLv3 (isnt it?)
The ZFS port to BSD was easy Solaris and BSD share the same low-level disk interface.
Linux is different. I've not looked at the Linux port - I avoid looking at Linux kernel code, it hurts my eyes - but I would guess it would need an adaption layer to run ZFS - assuming they've lifted the ZFS code 'asis' and are adapting the Linux low-level disk interface to allow ZFS to run.
Like all filesystems, the ZFS code ought to be kept separate from the main Linux kernel - which is pretty small anyways.
As the ZFS module is built as a kernel module ought to put a license firewall between it and the kernel.
"""GPL2 that force GPL2 onto derivative works have become a big obstacle to progress."""
Progress to whom? I think Linux is doing more than OK
"""The same goes for any other restrictive document such as the US constitution and magna carta"""
Not the same by a long shot, and careful what you wish for introducing changes here.
The IT industry is a legal minefield of moronic proportions, this is so because most people refuse to use the appendage known as head for anything else than carry it around, and because there are plenty of sharks willing to kidnap the industry for their benefit.
The GPL is designed with a goal, prevent the code from being made closed source or dependent on closed source.
Most tech companies idea of what the IT industry should be have more to do with seeing themselves as the "U.S. Robots and Mechanical Men" corporation, than players in a free market engaging competition.
The GPL as it is right now has proven to stop this from happening, provides a level-playing field and benefited everybody that plays by its rules. If you do not like the rules, go write your own kernel, with licenses and casinos and... (futurama reference)
Having said that, I want ZFS, so perhaps making a petition to Oracle could be an idea, after all they could ship it with their Oracle Linux too.
Not the same by a long shot, and careful what you wish for introducing changes here.
You miss the point. Both the US constitution and Magna Carta have already been partially or extensively modified by politicians. Documents that are far more significant to humanity than a poxy software license have, by common agreement, been modified. They are sacrosanct, but not unmodifiable.
So considering GPL to be unmodifiable is the height of conceit. Even politicians have managed to get their shit together more often than software devs. That's a pretty poor situation for the industry.
> considering GPL to be unmodifiable is the height of conceit
I don't think anyone has said that they consider that, have they? Of course the GPL is modifiable: we're currently on Version 3, already. What you *can't* do is go back and retro-actively modify Version 2. Re-licensing the kernel under a later version would require the assent of all the kernel contributors who hold a copyright. I'm unaware that there's an appetite to undertake that task.
".....and it's only the foamy mouthed zealots who care......" Er. no. It's the people wanting to use Ubuntu-based distros for commercial products that will care, as it opens them up to Oracle later coming along and either demanding a license fee for every instance, or - if they are a competitor to an Oracle product - being bankrupted in court by Oracle's legal arm. That is what Oracle tried with NetApp. Oracle certainly doesn't care about FreeBSD because it is not used in a commercial manner but by hobbyists. Personally, I don't see a problem if hobbyists want to download and install ZFS separately from a distro, but bundling it with the distro is leaving users open to potential liability issues. Myself, I would flatly refuse to use ZFS anyway because I consider it a crap product even before you get round to the licensing issue.
The good news is that all this does is make Red Hat an even better choice for server Linux.
Er. no. It's the people wanting to use Ubuntu-based distros for commercial products that will care, as it opens them up to Oracle later coming along and either demanding a license fee for every instance, or - if they are a competitor to an Oracle product - being bankrupted in court by Oracle's legal arm.
Sigh. Then that would be a matter for them to worry about, not a matter for Canonical and certainly not the FSF, etc. The rest of us mere mortals would simply like to be able to use a modern fs based on open source code in a convenient way without having to mess around with making one's own kernel modules (trivial though that may be).
And as others have pointed out here are plenty of distributions that are pushing out the closed source Nvidia drivers, and no one is suing anyone about that. The people getting hot under the collar about the addition of the open source ZFS are being mightily inconsistent.
Do you prefer Btrfs?
"....Then that would be a matter for them to worry about, not a matter for Canonical and certainly not the FSF...." Yes it would, seeing as the role of the FSF is to keep "Linux" free of legal/licensing/patent entrapments. If Oracle really had no interest in pursuing future licensing then they would have offered ZFS under the GPL years ago, but they have stayed with Sun's deliberately restrictive licensing instead. After seeing how Sun used the "WAFL-is-ZFS" argument to attack NetApp patents in court, anyone thinking Larry Ellison is just keeping ZFS under the CDDL because he is too lazy to swap it over to GPL is forgetting that Larry's one interest in buying Sun was to monetise as much of the corpse as he could.
"....And as others have pointed out here are plenty of distributions that are pushing out the closed source Nvidia drivers....." Which is totally different as nVidia has firstly stated it has no intent to pursue any licensing money, whereas Larry has, it was why he bought Sun. And secondly because nVidia still makes money off the consumer hardware sales of graphics cards that get put in consumer/hobbyist Linux systems, which is nVidia's prime market, whereas Larry makes no money off any commercial Linux system sales, indeed they are competitive to his own server and storage products. Indeed, nVidia needs to encourage as many consumer buyers as possible, both in Linux and Windows markets, as it faces the threat of Intel and AMD both having embedded graphics in their CPUs with enough capability for the majority of desktop users.
".....The people getting hot under the collar about the addition of the open source ZFS are being mightily inconsistent....." Maybe then you need to understand the difference between a commercial company and a specialist consumer company before you consider the views of others. Personally, as I stated before, if you want to load your consumer/hobbyist system with ZFS then it is your right, just do try and accept that those of us that use Linux in business are not going to be so happy at the idea of seeing a core distro encumbered with a patent timebomb.
".....Do you prefer Btrfs?" Yes, which is ironic considering it is an Oracle-developed product licensed under the GPL.
FreeBSD is used in a commercial manner. It powers the NetFlix cache network. It is the basis of the Islion platform.
From the point of view of some commercial FreeBSD groups, Linux is a hobbyist OS gone pro and FreeBSD is a commercial grade/quality OS. I agree with them.
From a historical point of view, FreeBSD came from the commercial versions of Unix through the rework done by Berkeley (the BSD in the name). Linux was written by college student and spread out among the the hobbyist on the Internet. It was years later that Linux became commercial. Remember, SCO sued IBM at the turn of century claiming that it took something like IBM to make the hobbyist OS known as Linux into the commercial/professional OS that Linux has become.
Remember, SCO sued IBM at the turn of century claiming that it took something like IBM to make the hobbyist OS known as Linux into the commercial/professional OS that Linux has become.
You're seriously trying to prove a point by referencing what SCO said?
::boggle::
Vic.
SFC's charge is brain-dead stupid. Canonical is in the right. Canonical's action would in NO WAY harm the openness of either linux or ZFS. The bottom line is that SFC is stirring up trouble over a fabricated "problem" where no problem of any kind in reality exists.
So Sun, who wrote ZFS, was bought by Oracle. They own the rights.
Linux was a competitor to Sun's Solaris.
When Sun OpenSourced ZFS, they explicitly said they didn't want it in Linux, so the deliberately picked a GPL incompatible license. That is what they said - it wasn't a mistake.
Since then, the licence has not been changed. So Oracle probably thinks the same way.
If Oracle dual license their copyright code, only then will ZFS not have problems in Linux (BSD is ok, btw, Sun chose the licence that way, and explained their thinking)
What Oracle has to do with it ? They can release their code under any license they want or do not release at all, end of the story.
Except it's not Oracle that would be distributing the combined/derivative work. It's Canonical.
By the SFC's reasoning (which, having read, I am inclined to agree with) Canonical would need to either infringe the GPLv2 by distributing the Linux kernel under CDDL, or infringe the CDDL by distributing ZFS under GPLv2.
Oracle don't like having their copyrights infringed.
Any hint what might SFC say about e.g. distributions making nvidia.ko available to users? Because if this module can be legally distributed (and it is closed source), so can be ZFS on Linux . Which is open source, although incompatible with GPL which means it cannot be merged into Linux sources (IMO good thing, as it would bifurcate OpenZFS and made ZFS filesystems incompatible between platforms). To me it seems that SFC are simply stirring the trouble.
Making a module available is not the same thing as being allowed to legally distribute it. If as you say this module can be legally distributed, I'm sure you will have no difficulty to point us to an official document from Nvidia legal department pertaining to this topic.
Like a lot of people on this forum, you need a better understanding of strong copyleft licenses and their purpose. I'm not saying this kind of license is good or bad, it's just that they serve a purpose and if we don't agree we should simply avoid them.
@AC yes I am pretty certain that distributing nvidia.ko is legal, and that is because I've read discussions on this topic and this is how I learned about EXPORT_SYMBOL_GPL . If you use a bit of google-fu you will find history of breakages in nvidia.ko , whenever kernel developers take pieces of kernel functionality away from non-GPL modules, i.e. wrap exported kernel symbol under EXPORT_SYMBOL_GPL. I think last time it happened in kernel version 4.2
You might also learn that ZFS on Linux is very careful not to use such symbols either, as it would prevent ZFS modules from loading.
I am certainly not going to hire a lawyer to give an opinion for you. Lawyers are expensive, if you have this kind of money to throw away, be my guest. I can give you some links if you are interested, though:
Ongoing dispute over the value of EXPORT_SYMBOL_GPL() function
Examining Loadable Kernel Modules Under the General Public License V2
Short version: kernel module is considered "derivative work" if it relies on any of Linux own interfaces. These are all marked EXPORT_SYMBOL_GPL . On the other hand, a work which only employs "standard interfaces", that is set of functions which can be reasonably provided by a kernel of any operating system, is not considered "derivative work". Such functions can be recognized in kernel sources by EXPORT_SYMBOL.
All Canonical has to do is invoke the "larger work" component of CDDL and distribute the zfs.ko source under CDDL. Linux stays GPL, kernel module sources stay CDDL and nobody's license is infringed. At worst, the SFC can claim that Canonical has to provide a public domain/BSD licensed shim as a bridge (which is how others have handled such 'incompatibilities' in the past). As someone who was around the first time the CDDL vs GPL holy war came up, it just sounds like the SFC is still trying to bury one license for it's favorite.
Side note: The whole argument is kinda moot for me since I use FreeBSD which has had ZFS for pretty much ever now. It's just funny to me that something named the "Software Freedom Conservancy" is trying to create arguments to erase a license they don't like. Nothing like watching someone argue about the "right kind" of freedom :)
"Oracle famously fought a protracted lawsuit with NetApp over ZFS and who can employ it and when."
Could there be a situation in the future where someone produces a product or appliance that uses Ubuntu (including ZFS), where this product sells well and makes lots of money for its manufacturer? If so, might Oracle sue them for use of ZFS?
Oracle does not and cannot sue anyone for using ZFS, because it was made open source under CDDL license which allows such things as FreeNAS or ZFS on Linux. If someone distributed Oracle's own closed source branch of ZFS, that would be entirely different story - but I imagine they guard this source just like they guard Oracle's.
Well OK I think I failed to make it clear - I do not have beef with GPL, in fact I am very grateful to Linux Foundation , FSF , Red Hat and others for allowing me to use Linux, git, GCC and other GPL tools and continuously improving them. But I am bothered by lawyers picking fights against another open source license, simply because it is more permissive than GPL, especially when (as it appears to me) they do not have a leg to stand on.
Sorry to bring you down from your high horses but what they actually feel is the need to tell you about their definition of free. As long as you're talking about your own work you can define free anyway you like but when you're using GPL software you should make an effort to understand and follow the letter and the spirit of its license. Alternatively, you can bitch or laugh at it and walk away.
So, we're speaking of whose graduation here ?
In that case, distributing ZFS with Ubuntu would clearly put them in violation of GPL license, unless they can extend the benefits of that particular licensing to each and every Linux distribution. They can enter a licensing or patent agreement but it must not be an exclusive one, something similar to what Samba devs did when they paid Microsoft for the permission to use some of the IP related to SMB protocols.
I wonder if someone with a better memory than me can correct me on this...
Didn't this argument already rage some years ago about non-GPL binary-only modules being loaded in to Linux, and Linus settled this by implementing the "tainted" bit... so that a user can optionally load in a binary mod and it would "taint" the kernel - so any stack traces would show that it wasn't "pure"
And then I thought there were much discussions and possibly court cases which decided that loading "binary" modules was not the same as merging the source code, so therefore the incompatible licensing didn't apply.
Was this something to do with nvidia's closed source modules?
I hope someone can correct me on this, it's all a bit hazy now, and I may have imagined some of it...
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
