FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS [LWN.net]

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 12, 2016 2:54 UTC (Tue) by pabs (subscriber, #43278) [Link]

Another announcement post:

https://sfconservancy.org/blog/2016/apr/11/fsf-zfs/

Stallman Is Wrong

Posted Apr 12, 2016 8:46 UTC (Tue) by linuxrocks123 (subscriber, #34648) [Link] (32 responses)

This will ultimately come down to what copyright law says, not what the license says. And the copyright issues are clear.

The ZFSOnLinux ZFS distribution is clearly not a derived work of Linux: it uses the Solaris kernel APIs. To make it work on Linux, the ZFSOnLinux guys wrote a "Solaris Porting Layer" which translates Solaris kernel API calls to Linux kernel API calls. The Solaris Porting Layer is licensed under the GPLv2, so even if it is a derived work of Linux, that's fine.

To use ZFS on Linux, then, you download the GPLv2-licensed SPL and the CDDL-licensed ZFS. Let's recap the license / derived work situation:

- ZFSOnLinux's ZFS distro can't be a derived work of Linux; it was written for Solaris and uses Solaris kernel APIs.
- ZFSOnLinux's Solaris Porting Layer translates Solaris to Linux kernel API calls. Since it's calling Linux kernel APIs, it's maybe (but maybe not) a derived work of Linux, but no one needs to care because the SPL is GPLv2 licensed.
- ZFSOnLinux's ZFS distro is almost certainly not a derived work of the Solaris Porting Layer because the APIs of the Solaris Porting Layer are the same as those of the Solaris kernel.
- The Solaris Porting Layer is maybe a derived work of Solaris. The answer to this question is the same as the answer to the question, "Is a clean-room reimplementation of the Java Standard Library APIs a derived work of the original implementation of the Java Standard Library?". Unless you're high on hallucinogens, the answer is clearly no, but apparently the CAFC is high on hallucinogens because they ruled "yes" in Oracle v. Google. Fortunately this is a moot point because Oracle has distributed the SPL under the GPLv2 in Oracle Enterprise Linux. Thus, Oracle is estopped from claiming a license violation over the SPL.

So basically it's pretty clear there are no prohibited derived work relations between any of these components, and distributing the source code is fine. If you built binaries out of these, you'd get the following:
- SPL kernel module binary licensed under GPLv2
- ZFS kernel module binary licensed under CDDL

The reason Canonical thinks this is okay is because the ZFS kernel module is sort of like the NVidia kernel module: it wasn't originally written against the Linux kernel APIs and uses a shim layer instead. I haven't looked at the code, but my initial impression is I'm inclined to agree.

Stallman Is Wrong

Posted Apr 12, 2016 9:23 UTC (Tue) by niner (guest, #26151) [Link] (21 responses)

You may be right, but you are also missing the point so much that I wonder if you've actually read the FSF's statement.

No one questions the legality of distributing ZFSOnLinux. It is clearly not a derived work of the Linux kernel. No one questions the legality of the Solaris Porting Layer as it does use the kernel's license.

What's in question is the legality of _distributing_ the combined work, let's call it "Linux with ZFS support". Because _that_ is clearly a derived work of both the Linux kernel and of ZFSOnLinux and the SPL and the kernel's GPL license requires such derived works to be distributed under the terms of the GPL.

The result is that downloading the packages from zfsonlinux.org, compiling them, installing them and using them is no issue at all. Yet including ZFS in a Linux distribution and distributing it with the Linux kernel is dubious from a legal standpoint. If you disagree, then please include in your statement why the precise points in the FSF's statement are wrong or do not apply.

Stallman Is [not] Wrong

Posted Apr 12, 2016 13:38 UTC (Tue) by mebrown (subscriber, #7960) [Link] (19 responses)

You've hit the nail on the head right here.

Neither thing is a derivative of the other. But when you combine them, you make a derived work of both. As long as you are not redistributing, this is fine. But as soon as somebody like Canonical decides to redistribute, they must comply with the license terms of both modules, which in this case are mutually incompatible (and makes redistribution violate one or both licenses)

Stallman Is [not] Wrong

Posted Apr 12, 2016 15:16 UTC (Tue) by butlerm (guest, #13312) [Link] (15 responses)

> But when you combine them, you make a derived work of both.

That isn't at all obvious. Placing two files on the same distribution media do not make the aggregation into some sort of derivative work. Rather, what they make is a collective work:

// A “collective work” is a work, such as a periodical issue, anthology, or encyclopedia, in which a number of contributions, constituting separate and independent works in themselves, are assembled into a collective whole. // (17 USC 101)

A derivative work, on the other hand, is a work that is "recast, transformed, or adapted" from a preexisting work, in such a way as to create an "original work of authorship".

// A “derivative work” is a work based upon one or more preexisting works, such as a translation, musical arrangement, dramatization, fictionalization, motion picture version, sound recording, art reproduction, abridgment, condensation, or any other form in which a work may be recast, transformed, or adapted. A work consisting of editorial revisions, annotations, elaborations, or other modifications which, as a whole, represent an original work of authorship, is a “derivative work”. //

Putting two files next to each other on the same medium doesn't amount to much of an adaptation.

Stallman Is [not] Wrong

Posted Apr 12, 2016 15:29 UTC (Tue) by k3ninho (subscriber, #50375) [Link] (11 responses)

>Putting two files next to each other on the same medium doesn't amount to much of an adaptation.

I'm no lawyer, and AFAIK there's not a recorded piece of case law but ... were we arguing in court, you make a copy of a file or a binary program and adapt it to the filesystem that it resides on. That requires a decent amount of know-how to enable the computer to match the byte lengths and binary layout plus metadata, and typically telling a computer to do something amounts to creativity or culpability in copyright law. Add in an installer and lawyers arguing that this is a substantial adaption can clearly claim the intent is there to aggregate and transform in such a way as to make it possible to use the two together.

K3n.

Stallman Is [not] Wrong

Posted Apr 12, 2016 17:20 UTC (Tue) by drag (guest, #31333) [Link] (10 responses)

You are making a mistake of trying to apply 'common sense' to copyright law. This is a huge mistake.

I really doubt having to have 'Linux know-how' in order to make ZFS work on Linux meets the criteria of 'derivative works'. This is really to the point you need a professional lawyer who specializes in copyright law to make any sense of this stuff.

Remember always that copyright is very arbitrary on many points. Copyright and other 'IP' laws exist for one purpose in the USA: "To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries."... which means that the law exists only for incentivizing new inventions.

Classic example is that when you buy a music cd you can play it for yourself or even for friends at a party. But if you play that music cd at a restaurant to entertain your guests then that is a copyright violation since it's a 'public performance' and you have to obtain a different type of copyright license.

Does it make sense? Not for any sort of 'natural rights' or 'protecting the artists work' standpoint. It only makes sense from 'let make it easier for copyright holders to demand money' standpoint; Which is the entire point of copyright.

Stallman Is [not] Wrong

Posted Apr 12, 2016 18:37 UTC (Tue) by fandingo (guest, #67019) [Link] (7 responses)

You're letting your ideology make a mess of your post.

> Copyright and other 'IP' laws exist for one purpose in the USA: "To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries."... which means that the law exists only for incentivizing new inventions.

> Does it make sense? Not for any sort of 'natural rights' or 'protecting the artists work' standpoint. It only makes sense from 'let make it easier for copyright holders to demand money' standpoint; Which is the entire point of copyright.

The first point is simplistically naive. Congratulations, you know how to quote the Constitution -- now it's time to catch up on nearly 250 years of case law and statute. The second statement oversimplifies an ideological opinion to shock. This is not at all what copyright law is about. Copyright law is about creating property rights and regulating that property. The owners are free to enter into contracts for valuable consideration to sell or lease their property just like they would with any other unrestricted, legal property. Your two statements are at loggerheads.

> Remember always that copyright is very arbitrary on many points.

Is it? I certainly don't think so. Perhaps it's that people want cookie-cutter solutions that fit into their preconceived notions and opinions, and they find the depth and sophistication of (copyright) law too filling for their intellectual appetites.

> You are making a mistake of trying to apply 'common sense' to copyright law. This is a huge mistake.

The one part with which I agree. Like most things, common sense is over-simplistic and happily ignores domain-specific knowledge in the interests of dumb analogies (typically of the automobile variety) and gut feelings. The law is complex, and complex things require lots of study, attention to detail, and careful understanding of side interactions. Common sense couldn't help you learn to write a Fast Fourier Transform algorithm, so why could it help you understand the intricacies of copyright law?

Stallman Is [not] Wrong

Posted Apr 12, 2016 19:44 UTC (Tue) by drag (guest, #31333) [Link] (3 responses)

> Copyright law is about creating property rights and regulating that property. The owners are free to enter into contracts for valuable consideration to sell or lease their property just like they would with any other unrestricted, legal property. Your two statements are at loggerheads.

No they are not. For copyright law to function it has to be at odds with all other forms of property rights.

Property rights exist for a purpose. That purpose is to enable society to have the ability to preemptively resolve conflicts on scarce/finit/conflicting use of resources. While legal details about property rights themselves may be somewhat culturally-specific they all deal with trying to resolve this same problem.

Whether or not property rights are codified by state law is actually irrelevant. It is something that grew out of common law and even if you do not understand the concept you are still going to need to learn and adhere to property rights in order to get along with other people in society. It's something that just naturally needs to occur as conflicts over scarce resources is a constant problem in all human societies. It's part of how humans interact with one another. You can go to the most primitive tribe or the most advanced socialist country while the property rights themselves may have significant variations they all focus on solving the same sets of problems.

Notice how property rights do not apply to resources that are NOT scarce. So-called public goods are not 'owned' by anybody.

Copyright law does not establish controls over scarce resources. Things like 'ideas' and 'poetry' and 'source code' are metaphysical concepts meant to encapsulate patterns of thoughts and as such are not scarce resources. ie: If I have a idea on how to make a wheel and you copy that idea I am not denied the usefulness of my idea. Once a idea or story or concept is created it can be copied infinite amount of times with no loss to anybody.

Thus copyright law itself is separate and distinct set of laws from property rights. The purpose is different, how the law works is different, and how it applies to day to day life is different. What copyright creates is a monopoly over a particular idea/story/etc for purposes of creating a 'market of ideas'. How this is accomplished is, essentially, a violation of real/traditional/common law property rights. It's a market restriction to reduce or eliminate competition in exchange for the act of 'creation'. That is how it works, that is what it does.

For example: You have a radio that picks up broadcast transmissions. You record a tape of a broadcast with particularly interesting song. You listen to it, and learn the song, and then you perform the song on stage. This is a copyright violation. You didn't 'steal' the song from anybody (except in a very abstract sense).

The reason this is a infringement on physical property rights is because it establishes controls over your radio, your tape recorder, your guitar, and even your physical body to somebody else. Now you cannot act for your own good without consent from a third party. Nothing that you will ever do with that song or idea will ever take a single thing from anybody else... EXCEPT for their ability to market it to the people that attended your concert.

This is very different then the sort of restrictions you get from having a car or somebody who trespasses on your land. Their use of your property denies your access to it. If somebody takes your car this means you have lost access to that car. If somebody copies your source code from github it does not prevent you from using that software.

With the GPL license you are taking advantage of market restrictions against competition that copyright introduces in order to try to extract payments from third parties in the form of source code redistribution and access to their improvements. This is 'wealth' that you are using copyright to try to extract in exchange for mostly relinquishing the controls over other people that the state has granted you through copyright laws. The payments in this case are not made in the form of cash money, but in the form of source code.

Saying that copyright as 'property rights' is just a really bad analogy that essentially acts as a layer of abstraction that hides how copyrights laws actually function.

Stallman Is [not] Wrong

Posted Apr 12, 2016 20:29 UTC (Tue) by fandingo (guest, #67019) [Link] (2 responses)

Go make this diatribe to an attorney and watch her expression. I imagine her eyes will grow wide and start darting to and fro as she desperately searches for an exit from the conversation.

There is nothing in law that distinguishes rights based on scarcity. I don't have the slightest idea where you're getting this idea. Wait, yeah, I do: the never-wrong common sense. (Weren't you criticizing this approach to copyright law in your previous post???) Ignoring that for the moment, of course there is scarcity in intellectual property. Trademarks, patents, and copyrights are all built on uniqueness, which is the penultimate form of scarcity. Sure, reproduction doesn't typically cost much, but that's not where the *value* originates with intellectual property. The scarcity is coming up with the ideas that customers want to purchase, lease, or license.

> What copyright creates is a monopoly over a particular idea/story/etc for purposes of creating a 'market of ideas'.

What property law creates is a monopoly over a particular item for purposes of creating a `market of items.' See, I can make up gobbledygook, too. Claiming that copyright exists to create "market" is a bridge too far. Again, this conflicts with your earlier quoting of the Constitution. Copyright law exists, so that creators retain exclusive control over their works, reproductions, and covered derivatives for a finite period of time and allows them to dispose of the property in the lawful manner they find most agreeable. That's it. Lots of copyrighted works are created without any intention of entering anything called a "market." (My diary is copyrighted, and it will never be part of any market -- financial or "of ideas.")

> For example: You have a radio that picks up broadcast transmissions. You record a tape of a broadcast with particularly interesting song. You listen to it, and learn the song, and then you perform the song on stage. This is a copyright violation. You didn't 'steal' the song from anybody (except in a very abstract sense).

Of course not, you infringed their copyright -- at least twice... Not sure what incredibly obvious point you're trying to make. Who is still talking about copyright infringement in terms of theft?

> Now you cannot act for your own good without consent from a third party.

Ohh, please. Who entered into that license on her own volition? Who chooses to live in a country where one is subject to all of its laws, including bog-standard copyright laws? If you find yourself a slave in this corporate-government machina, blame yourself.

Was this whole argument made using your common sense, or did you consult some legal resources? Your obsession with tying intellectual property to (runaway) capitalism is ludicrous.

Stallman Is [not] Wrong

Posted Apr 12, 2016 23:41 UTC (Tue) by drag (guest, #31333) [Link]

> Go make this diatribe to an attorney and watch her expression. I imagine her eyes will grow wide and start darting to and fro as she desperately searches for an exit from the conversation.

I wouldn't be surprised. Certainly my opinions wouldn't make one bit of difference in a lawsuit.

> There is nothing in law that distinguishes rights based on scarcity. I don't have the slightest idea where you're getting this idea. Wait, yeah, I do: the never-wrong common sense. (Weren't you criticizing this approach to copyright law in your previous post???)

There is different types of law. There are different ways they are originate.

Property rights originated via common law, which was later codified by the State. Property rights of one form or another are fairly universal to human societies. They need ways to divide up scarce resources between individual actors. Not all law originates via government edict. Marriage is another example of common law that was later regulated by States.

Copyright restrictions are much more modern and the ones we operate under in the USA originated with English censorship laws. I could go on for a few pages about this, but it's not really that important. In the USA copyright law is instituted to try to incentivize the publishing of 'creative works'. That is it's purpose. Give up some individual liberty in exchange for rewarding publishers. The hope is that would cause more things to be published.

Regardless '250 years of case law' doesn't change the nature of copyright or it's purpose or why it's in place in the first place. In fact those 200+ years of case law only exist to refine and reinforce the original purpose of copyright.

> Copyright law exists, so that creators retain exclusive control over their works, reproductions, and covered derivatives for a finite period of time and allows them to dispose of the property in the lawful manner they find most agreeable. That's it. Lots of copyrighted works are created without any intention of entering anything called a "market."

This does not support your notion of 'copyright is a form property rights'.

Copyright is not necessary for creators to retain 'exclusive control over their works'. The only way they retain exclusive control, whether copyright exists or not, is to simply not publish their works. Once published the author no longer has exclusive control. They only have the ability to sue people that violate his copyright in a manner that is consistent with copyright law.

The nature of the original discussion should make this specific point painfully obvious... that is GPL being limited by the State's definition of 'derivative works'. So on that point you are very easily wrong.

Also all things covered by copyright are covered regardless of whether or not you want it or not. That is simply a artifact of the law that is in place in order to stop the need to for the government to maintain a large copyright registry office. It also establishes a logical baseline for the expiration of copyright restrictions (again without forcing anybody to maintain a database of copyrights. In a dispute over date somebody needs to just produce a copy of the work with a particular date to establish a baseline)

Incidentally this points to another major difference between copyright restrictions and property rights... Property rights do not expire.

> Of course not, you infringed their copyright -- at least twice... Not sure what incredibly obvious point you're trying to make.

That 'Intellectual Property' concept is confused and incorrect. That copyright has nothing to do with property rights, but in fact are infringements of property rights. Copyright exist to it illegal to perform certain actions with a person's own property and thus cannot be a form of property rights. Copyright and property rights naturally conflict with one another.

It is the point the one you took issue with with your first post.

Many State laws conflict with property rights. The logic is always the same: lose a bit of liberty for the benefit of society in the long run.

> Ohh, please. Who entered into that license on her own volition? Who chooses to live in a country where one is subject to all of its laws, including bog-standard copyright laws? If you find yourself a slave in this corporate-government machina, blame yourself.

Whether or not copyright is tyrannical or not is completely besides the point. It's fully possible to agree with everything I said while believing that copyright is valuable, desirable, and that life would be worse without these restrictions.

> Was this whole argument made using your common sense, or did you consult some legal resources?

Lots of resources over lots of years. Not a lawyer. It's a form of entertainment, ultimately.

> Your obsession with tying intellectual property to (runaway) capitalism is ludicrous.

You seem to be injecting something of your own in there.

Stallman Is [not] Wrong

Posted Apr 15, 2016 20:06 UTC (Fri) by nix (subscriber, #2304) [Link]

Who is still talking about copyright infringement in terms of theft?

A great many copyright maximalists, that's who. Some of them even know they're lying.

Stallman Is [not] Wrong

Posted Apr 12, 2016 21:12 UTC (Tue) by zlynx (guest, #2285) [Link] (2 responses)

> The first point is simplistically naive. Congratulations, you know how to quote the Constitution -- now it's time to catch up on nearly 250 years of case law and statute.

There are many people including lawyers and judges that believe that if it isn't in the text of the Constitution, then it does not apply. I believe this is called "strict construction."

An appeals court judge is free to ignore all case law and statute when looking at the Constitution because it is the ultimate authority.

Stallman Is [not] Wrong

Posted Apr 12, 2016 21:35 UTC (Tue) by fandingo (guest, #67019) [Link]

> An appeals court judge is free to ignore all case law and statute when looking at the Constitution because it is the ultimate authority.

ABSOLUTELY NOT. A federal appeals court is bound by two levels of precedent. First, that court's own decisions are binding precedent (insofar as they don't contradict with #2). Second, the Supreme Court's decisions bind ALL courts, including the Supreme Court. It's called *stare decisis*, and it's a cornerstone of the American Judiciary. It is a tremendously important part of keeping justice mostly uniform and consistent over time and across the country. Additionally, a federal appeals court is not the ultimate authority. That's the aforementioned Supreme Court. (There are some odd-ball states like New York that call their highest court the Court of Appeals, but federal courts have exclusive jurisdiction over copyright law.)

> There are many people including lawyers and judges that believe that if it isn't in the text of the Constitution, then it does not apply. I believe this is called "strict construction."

Strict constructionism isn't really an ideology; it's more of a dog-whistle used by those who want to caricature their opponents. The scholars who subscribe to similar ideas call themselves textualists or originalists, and there are important differences.

Stallman Is [not] Wrong

Posted Apr 12, 2016 23:09 UTC (Tue) by rahvin (guest, #16953) [Link]

> There are many people including lawyers and judges that believe that if it isn't in the text of the Constitution, then it does not apply. I believe this is called "strict construction." An appeals court judge is free to ignore all case law and statute when looking at the Constitution because it is the ultimate authority.

Wow. Just wow. That is so many ways of wrong it's not even funny. Supreme court precedent on copyright is well established and the chance of any radical change is so vanishingly small that it's practically non-existent.

Stallman Is [not] Wrong

Posted Apr 13, 2016 13:50 UTC (Wed) by k3ninho (subscriber, #50375) [Link] (1 responses)

>You are making a mistake of trying to apply 'common sense' to copyright law. This is a huge mistake.
We have a big disconnect. Your other comments indicate a running assumption that USA law is universal, and the USA Constitution is universal -- I don't live under that setup.

Most importantly, I think you've completely missed the point of my comment. I wrote a hypothetical response, not 'common sense' but instead what might convince judiciary and/or my peers. Thank you for sharing your opinion.

As an aside, I don't actually believe that there's enough 'common sense' that it's helpful to assume a pool of shared attitudes and experiences and to talk about that pool. This internet-based society is too diverse and has people with lived-experience too specialised for there to be much common overlap (there's even that saying: 'common sense aint so common').

K3n.

Stallman Is [not] Wrong

Posted Apr 13, 2016 17:19 UTC (Wed) by Wol (subscriber, #4433) [Link]

> As an aside, I don't actually believe that there's enough 'common sense' that it's helpful to assume a pool of shared attitudes and experiences and to talk about that pool. This internet-based society is too diverse and has people with lived-experience too specialised for there to be much common overlap (there's even that saying: 'common sense aint so common').

And, as I often put it with Science, "If it's 'common sense', it's almost certainly wrong".

The EASIEST way to get a *good* grade when studying relativity at Uni, is to put aside your common sense and treat it as a pure maths problem. The reason people find Special Relativity hard, is because common sense screams "this can't be right" when, actually, it is. (The reason people find General Relativity hard is because, well, it *is* hard. :-)

Cheers,
Wol

Stallman Is [not] Wrong

Posted Apr 12, 2016 15:54 UTC (Tue) by pboddie (guest, #50784) [Link] (2 responses)

But although the GPL deals with topics like distribution, it is the "live" work that governs the licence's considerations. So, upon combining and running Linux and, in this case, some ZFS implementation or other, it is this whole "live" work and its properties that the GPL seeks to regulate. This is because those terms were put on the GPL-licensed parts (in order to uphold the four freedoms) and are intended to ensure that other things in the resulting system also do so.

The only "get out" here is whether a copyright regime regards the two components as completely separate things and that no derivative has been made, even though the ZFS component will be reliant on Linux to function. And since the FSF is saying that you may not combine GPL-licensed things with GPL-incompatible things, that "get out" is strongly confined to a judge saying that "people shouldn't pay attention to those guys, it's all OK because nothing gets mixed up" which is pretty questionable.

And if they were to rule that "stuff does get mixed up but it's still OK" or even "no-one cares about all that, I still say it's OK" then they would cause a great deal of trouble for copyright in general, so that would be rather unlikely as well.

Stallman Is [not] Wrong

Posted Apr 12, 2016 17:32 UTC (Tue) by drag (guest, #31333) [Link] (1 responses)

> The only "get out" here is whether a copyright regime regards the two components as completely separate things and that no derivative has been made,

That's not the 'get out'. GPL is a copyright license and as such it is limited by scope of copyright. The scope of copyright restrictions is limited by 'derivative works' as legally defined by the legislation and in case history.

The GPL nor FSF does not get to define what is and what is not derivative works. Anything that the FSF says about what is and what is not derivative works can be comfortably ignored, unless it happens to be correct. And 'correct' is defined by 'whatever a Judge happens to decide is relevant in a lawsuit'

May want to try to decipher this stuff:

http://www.law.washington.edu/lta/swp/Law/derivative.html

It contains fun nuggets like:

> As usual, the answer in this example depends on questions of copyrightability, incorporation, substantial similarity, market demand, and public policy. As a general matter, in the case of inheritance in object oriented systems, a subclass trivially incorporates (and is substantially similar) to its superclass. The analysis should then turn, to a larger degree, on questions of copyrightability of the superclass, market demand, and public policy. For example, if the class B is highly generic (as in Java's Object class), it is unlikely that it would even be considered copyrightable subject matter under the 102(b) bar. On the other hand, if B is highly specialized—as are classes that we find on the fringes or leaves of a class hierarchy—there may be a stronger argument that they contain substantial copyrightable subject matter. This argues, then, for a finding that D is a derivative of B. Even here, though, a market demand analysis might argue against a finding of derivation. Class D will not supplant demand for class B, simply because it makes no sense to deploy D in the absence of B.

Stallman Is [not] Wrong

Posted Apr 12, 2016 20:07 UTC (Tue) by mattheww (guest, #108205) [Link]

Copyright law allows copyright owners to give limited permission to distribute their copyrighted works.

A copyright owner could give permission to distribute their work in a red box but not a green one, even though they have no special rights over green boxes.

A copyright owner could give permission to distribute their work together with a recording of Beethoven's 5th symphony but not a recording of the 1812 overture, even though they have no special rights over the 1812 overture.

A hypothetical super-strong copyleft licence could give permission to distribute a work only if all other works on the same medium of exchange are licensed under the same super-strong terms, _even if_ the relationship of those works is "mere aggregation".

So I think anyone trying to determine what GPLv2 allows when someone distributes a GPLed work together with other stuff should be interpreting the terms "part of a whole which is a work based on the Program" and "mere aggregation", and ignoring "derivative work".

(It might turn out that what the GPL effectively covers is very similar to what some country's copyright law determines to be "derivative works", but it doesn't matter whether it does or not.)

Stallman Is [not] Wrong

Posted Apr 12, 2016 18:50 UTC (Tue) by gwolf (subscriber, #14632) [Link] (2 responses)

If this was true, all of the distributions would have a hard(er) time getting basically any collection of differently-licensed software to be legally distributed. As an example, whenever we have GPLed code in Debian that requires OpenSSL, we have to talk the author into licensing it — For further discussion, see as examples:

• https://github.com/OfflineIMAP/offlineimap/issues/104 (OpenSSL and GPL incompatibility)
• https://people.gnome.org/~markmc/openssl-and-the-gpl.html (The OpenSSL license and the GPL)
• https://opensource.stackexchange.com/questions/2233/gpl-v... (GPL V3 with OpenSSL Exception)

If Linux and ZFSOnLinux are shipped in the same media, they remain two clearly differenced and separated products. Their distribution remains legal. Otherwise, we would be unable to ship OpenSSL and any GPLed software in the same media.

Stallman Is [not] Wrong

Posted Apr 12, 2016 23:33 UTC (Tue) by rahvin (guest, #16953) [Link] (1 responses)

Whether or not it's legal aggregation or illegal combination would be for a court to determine based on the laws of the jurisdiction in which the case is brought. Anyone asserting that it's one or the other is doing so at risk as there are no real solid precedents in the case law for this situation. The point a lot of people have been trying to make is that this is a risk and if one of the copyright holders doesn't like it they can sue and make things very unpleasant for the person that distributed the combined product.

Just as an aside, common knowledge was that API's could not be copyrighted, right up until Google lost and the appeals court declared API's eligible for copyright. Just something to thing about.

Stallman Is [not] Wrong

Posted Apr 14, 2016 7:19 UTC (Thu) by linuxrocks123 (subscriber, #34648) [Link]

The Oracle v. Google decision is not binding except in the Federal Circuit when the Federal Circuit is interpreting Ninth Circuit law. It is therefore of limited impact, and the Ninth Circuit can overrule the decision later since the Federal Circuit is bound by the case law of the home circuit.

The decision is also just plain wrong, so the Supreme Court will have to overrule it someday if the Ninth Circuit doesn't. No surprise there; the Supreme Court is always having to clean up after the Federal Circuit's incompetence.

Stallman Is Wrong

Posted Apr 12, 2016 17:23 UTC (Tue) by linuxrocks123 (subscriber, #34648) [Link]

I'm not missing the point; you failed to read my entire comment. The last part talks about why Canonical thinks distributing the binary modules is okay.

Stallman Is Wrong

Posted Apr 12, 2016 11:20 UTC (Tue) by mjg59 (subscriber, #23239) [Link] (6 responses)

> The ZFSOnLinux ZFS distribution is clearly not a derived work of Linux: it uses the Solaris kernel APIs.

I'm not going to comment on any other aspect of this, but this is factually incorrect - ZFSOnLinux makes direct calls to Linux-specific APIs.

Stallman Is Wrong

Posted Apr 12, 2016 11:25 UTC (Tue) by kiko (guest, #69905) [Link] (2 responses)

So what? How does that imply it's a derived work?

Stallman Is Wrong

Posted Apr 12, 2016 11:34 UTC (Tue) by mjg59 (subscriber, #23239) [Link]

I'm not in a position to provide a factual assertion as to whether or not it's a derived work. I was pointing out that any claim that ZFSOnLinux is only interacting with Linux via the Solaris Portability Layer is untrue.

Stallman Is Wrong

Posted Apr 18, 2016 0:10 UTC (Mon) by ksandstr (guest, #60862) [Link]

(IANAL, obviously.)

By the combination of ZFS and its Linux shim being an egg to the chicken, regardless of how ZFS was originated. The same was true of XFS.

Stallman Is Wrong

Posted Apr 12, 2016 17:33 UTC (Tue) by linuxrocks123 (subscriber, #34648) [Link] (2 responses)

After looking at the code, I was able to find at least one instance of the ZFS module calling a Linux kernel API function directly (find_lock_page). Whether that changes the analysis depends on how frequently this occurs, to what extent, and how Linux-specific the API calls are. What do you know about this?

Stallman Is Wrong

Posted Apr 13, 2016 16:24 UTC (Wed) by jospoortvliet (guest, #33164) [Link] (1 responses)

Pure speculation here but - any chance that that function is either public API or not exported as GPL_only? If so, I guess it doesn't matter much.

Stallman Is Wrong

Posted Apr 13, 2016 17:03 UTC (Wed) by mjg59 (subscriber, #23239) [Link]

Can you elaborate on what "public API" means in this context, and why the GPL_ONLY nature of an export matters?

Stallman Is Wrong

Posted Apr 12, 2016 12:02 UTC (Tue) by vonbrand (guest, #4458) [Link] (2 responses)

The Solaris Porting Layer, in turn, is presumably a derived work from Solaris, and thus must be distributed under CDDL. As the licenses are incompatible, the shim can't be distributed at all.

Stallman Is Wrong

Posted Apr 12, 2016 12:06 UTC (Tue) by Funcan (guest, #44209) [Link]

No, the license holders (Oracle) can decide to relicense it, which they did.

Stallman Is Wrong

Posted Apr 13, 2016 0:40 UTC (Wed) by kiko (guest, #69905) [Link]

The SPL is licensed under the GPLv2 (or later): https://github.com/zfsonlinux/spl/blob/master/COPYING

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 12, 2016 9:07 UTC (Tue) by gowen (guest, #23914) [Link] (10 responses)

Is there absolutely any precedence or case law - in any legal domain - that supports the idea that "linking creates a derivative work"? Or is it a legal opinion obtained by the FSF that they bat around as if it were an incontrovertible fact?

ProTip: If someone tells you "X is prohibited by law" but doesn't tell you the jurisdiction in which such a prohibition holds, they are attempting to sell you a bill of goods.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 12, 2016 12:06 UTC (Tue) by Wol (subscriber, #4433) [Link] (3 responses)

The jurisdiction is "International Law" - aka in this case Berne - so it holds for pretty much all the civilised world. Whether or not the world is civilised is a debate best not held here ... :-)

Cheers,
Wol

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 12, 2016 13:19 UTC (Tue) by gowen (guest, #23914) [Link]

Well, that's not right for a start.

Berne has codified protection for derivative works, but no criteria for what constitutes one, and no WIPO cases that rule on the matter. The US Copyright Act does give a definition (untested, to my knowledge, w.r.t. static-linking-as-derivation) but this is in no way binding on non-US signatories to Berne. Similarly, US rulings on parody, fair-use etc., are valid only in the US and are not precedent elsewhere.

This has been a fake-internet-lawyer presentation.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 12, 2016 13:22 UTC (Tue) by gowen (guest, #23914) [Link] (1 responses)

Further to the above, here's what Berne has to say about Derivative Works:

"(3) Translations, adaptations, arrangements of music and other alterations of a literary or artistic work shall be protected as original works without prejudice to the copyright in the original work."

That's it. If you think that clears anything up on this issue, you have a sharper insight than mine.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 12, 2016 13:44 UTC (Tue) by Wol (subscriber, #4433) [Link]

As I read that, it seems pretty clear to me.

It says that if I adapt your work, then I have a new copyright in my work. You can't come to me and say that you want a share in my (new) work.

It does NOT preclude you, however, from claiming that my work is a derivative of yours and that I therefore owe you one for it ...

(In other words, you have no claim against my work. You may, however, have a claim against me.)

Cheers,
Wol

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 16, 2016 20:40 UTC (Sat) by davidstrauss (subscriber, #85867) [Link] (5 responses)

Thank you for stating this so well. I'm a big fan of LWN, but they frequently present the "linking = derivative" legal theories of the FSF, SFLC, and similar advocacy organizations as if they're well-supported by case law (which is quite relevant in common-law jurisdictions like the United States).

I personally find "linking = derivative" to be a frightening theory because it wouldn't just hold for copyleft works; it would mean linking to proprietary code makes work a derivative of that, too. In response to this concern, I often hear "But, the GPL defines that linking creates a derivative work, while licenses for the Windows API don't make such claims." At least in the US, what is a derivative work is a concern of the courts, not the text of the license. A license can grant rights *if* a work is a derivative, but it cannot determine whether it is in the first place. If a license could do that, the results would be absurd because license authors would have incentive to define derivative works as broadly as possible.

I support the work of organizations like the FSF and count many of their employees and supporters as my friends. However, there's a goal underpinning their legal theories that, combined with the paucity of case law, means their statements should be treated with more skepticism in neutral reporting.

I'm not a lawyer, but I've also seen the skepticism I've stated above (in terms of linking creating derivatives) expressed by Google's legal team. Lawyers works for whoever hires them, and their theories suit their clients. This is true for both the FSF and Google.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 17, 2016 9:45 UTC (Sun) by SiB (subscriber, #4048) [Link] (1 responses)

I do not think the FSF claims that linking makes a work derivative. How I understood it, what they say is that when a work builds upon another work, whether it is derivative does _not_ depend if it is linked statically or dynamically, linked by the user or the distributor, if it is distributed seperately or not. The new work may be derivative even before it is linked to - and without including - the original work.

License terms can depend on linking and distribution methods.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 17, 2016 18:19 UTC (Sun) by davidstrauss (subscriber, #85867) [Link]

"It has always been the FSF's position that dynamically linking applications to libraries creates a single work derived from both the library code and the application code." [1] Considering that dynamic linking is the less substantial of the two types (versus static) and that copyright has no sense of a "library" versus an "application," the implication is that linking creates derivative works.

The FSF's own FAQ on the GPLv2 also takes this position by stating that linking to a GPL'd work requires making the entire work GPL'd [2]. This would only be true if it created a derivative work. This is not just my interpretation of that FAQ item: "The Free Software Foundation GPLv2 FAQ is unequivocal that all forms of linking against a GPL-licensed binary create a derivative work, so that the terms of the GPL apply to all of the linked modules; many developers agree with this interpretation." [3]

[1] https://www.gnu.org/licenses/lgpl-java.en.html
[2] https://www.gnu.org/licenses/old-licenses/gpl-2.0-faq.htm...
[3] https://lwn.net/Articles/548216/

linking = derivative

Posted Apr 17, 2016 14:51 UTC (Sun) by corbet (editor, #1) [Link] (2 responses)

So I'm curious: where have we presented "linking = derivative" as established case law? Inasmuch as LWN has beliefs, that's not one of them...

linking = derivative

Posted Apr 17, 2016 18:47 UTC (Sun) by davidstrauss (subscriber, #85867) [Link] (1 responses)

Indeed, LWN is not *explicitly* presenting it as established case law. However, LWN does imply it is by presenting the FSF's position uncritically when the FSF is asserting its position in jurisdictions like the United States or the United Kingdom -- places where case law is the only thing that could make their position on linking more than a (disputed) legal theory. Absent that case law, the FSF's positions are ones of advocacy, not neutral "explainers" like on SCOTUSblog.

For (1), in the case of this story, the introduction saying "how it is incompatible with the CDDL" implies that the incompatibility is well-established/undisputed and that the FSF is either providing an explanation of the agreed-upon reason or, at least, its interpretation of the reason. Jumping right to "how" begs the question (in the logical sense [1]) by assuming the "what" of "linking = derivative."

For (2), at least Canonical seems to think they're in the right, given that they're the ones distributing the GPL and CDDL components together. They probably asked their lawyers first. While I believe they are using some tricks to perform the linking on the installed machines (rather than before distribution) and by using a shim, the FSF thinks at least the latter isn't good enough.

It's obviously not good journalism to "present the controversy" when there isn't one, but I think there is one here. Part of why it may seem to not be controversial is an issue in terms of loudness from each side. Advocates of "linking = derivative" run multiple organizations dedicated to popularizing that theory. Organizations believing "linking != derivative" usually just quietly go about their work. That doesn't mean the latter group doesn't have a thoughtful position.

[1] https://en.wikipedia.org/wiki/Begging_the_question

linking = derivative

Posted Apr 17, 2016 23:38 UTC (Sun) by paulj (subscriber, #341) [Link]

My experience of legal advice on stuff in this area - e.g. from valley tech corp legal counsel - is that the lawyers do not care much about the software mechanics "linking". They care about a much more high-level analysis of the issues. Not being a lawyer, I can't really say how it should work, but I do know that when I hear technical people argue software derivation in terms of mechanics that that just doesn't tally with how the actual lawyers I've dealt would reason about it.

Also, my experience is that the tech corp counsel are even more hard-core on the derivation issue than the FSF... Maybe that was just that one corporation, but when I see how far other corporates will go with GPL-fear, I wonder if it isn't the same at those others too.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 12, 2016 16:16 UTC (Tue) by jhoblitt (subscriber, #77733) [Link] (1 responses)

I feel that folks are losing forest for the trees while lawyer over technical copyright issues. IMHO - the core issue here is that we have a party (Oracle) that by all appearances has intentionally created legal ambiguities for any third party that may want to redistribute their software, presumably for some sort of marketplace advantage. They have not responded to multiple years worth of requests to resolve the licensing situation. At what point do we (the community) decide that a a party is willfully acting in bad faith and we don't want what they are selling? How much quality code could have been constructed with the amount of energy that has been spent debating this side show?

The software in question is not a critical missing component of a server software stack nor does it contribute to solving the challenging issues we're facing today trying to scaling storage infrastructure. Move along, nothing worth seeing here...

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 13, 2016 2:53 UTC (Wed) by pabs (subscriber, #43278) [Link]

Sun created the situation, Oracle is merely continuing it.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 12, 2016 19:29 UTC (Tue) by JoeBuck (subscriber, #2330) [Link] (11 responses)

As I am not a lawyer I will not offer an opinion as to who is right; I can see some merit in both the RMS argument and the Canonical argument. But I think it's fair to ask Oracle to clean up the mess by simply putting out a statement saying that it is permissible to use ZFS under the GPL as well as the CDDL, and think it would be wise for Oracle to go along with this. That eliminates the argument entirely. They no longer have much of reason to want to protect Solaris; their main business these days is selling database software that runs either on Linux or Windows.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 12, 2016 20:50 UTC (Tue) by dlang (guest, #313) [Link] (6 responses)

The fact that Oracle is not willing to do this should make everyone who wants to use ZFS concerned.

Remember, this is the company that waited years to file suit against Google for Android using Java function names.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 12, 2016 23:40 UTC (Tue) by rahvin (guest, #16953) [Link] (5 responses)

There is also significant discussion (available on the internet) from former SUN employees that were party to the original licensing discussion that have said the CDDL was made deliberately incompatible with the GPL. These testimonies would likely be brought up in any court case. Along with Oracle being what they are, this original knowledge and other factors I wouldn't want to distribute a combined product, as I'm willing to bet the copyright is registered and you'd be up for statutory damages of $150k per copy.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 13, 2016 8:51 UTC (Wed) by paulj (subscriber, #341) [Link] (4 responses)

There is *one* statement from *one* Sun employee. Another Sun employee - in the same Debian workshop - disagreed with her on that in comments almost immediately after.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 13, 2016 11:52 UTC (Wed) by pizza (subscriber, #46) [Link] (3 responses)

> There is *one* statement from *one* Sun employee.

When that *one* statement was made very publicly, and that *one* employee was Sun's corporate Open Source Officer (who oversaw all of Sun's F/OSS efforts and was attributed as one of the primary authors of the CDDL), that *one* statement carries quite a lot of weight.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 13, 2016 14:46 UTC (Wed) by nye (guest, #51576) [Link] (1 responses)

In fact, Sun's Open Source Officer at the time (Simon Phipps) was the one who *disagreed*.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 13, 2016 15:14 UTC (Wed) by pizza (subscriber, #46) [Link]

You're correct, I didn't check that reference.

However, at the time of that "statement", Mr. Phipps introduced Ms. Cooper as "the one who actually wrote the CDDL". So her words carry quite a bit of weight.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 13, 2016 15:22 UTC (Wed) by paulj (subscriber, #341) [Link]

Cooper made a brief comment. Simon expanded on it. Cooper said the CDDL was selected because it was GPL-incompatible because of the engineers. Simon said the engineers didn't want GPL but wanted the BSD licence - and would have revolted if GPL was selected.

I am confident Simon's characterisation is the more accurate, cause my mentor in Sun was a senior engineer involved in those discussions, and what I heard from my mentor tallies with what Simon has said in public. Quite a few in Sun engineering wanted *BSD* - Sun engineering had a long-standing, strong BSD bent - and an associated cultural antipathy toward the GPL.

Danese's characterisation that CDDL was selected to be GPL-incompatible because of the engineers can not be correct, if engineering wanted BSD.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 13, 2016 8:41 UTC (Wed) by lmb (subscriber, #39048) [Link] (2 responses)

But could Oracle do this? How many copyright holders would they need to contact to relicense, even if they wanted to?

And - the "new" ZFS on Solaris is already quite ahead with features and again closed, or am I misinformed?

I'm seriously not getting why the energy in trying to port something that was meant to be Linux-unfriendly isn't used for improving those codebases that are.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 13, 2016 18:37 UTC (Wed) by JoeBuck (subscriber, #2330) [Link] (1 responses)

The Oracle corporation owns the copyright, 100%. The contributors were Sun employees, and it was a work made for hire, and Oracle bought Sun and all of its patents, copyrights, and trademarks.

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 13, 2016 18:49 UTC (Wed) by mjg59 (subscriber, #23239) [Link]

Oracle owns the copyright on ZFS, but there have been significant contributions to ZFSonLinux from non-Oracle employees. Oracle would not be able to relicense most of that code without contacting the copyright holders.

(Well, that's not entirely true. Oracle could release a new version of CDDL that permitted relicensing of CDDLed material under GPL-compatible terms, and with the exception of two small files that explicitly pin themselves to CDDL 1.0 the entire problem would then be solved)

FSF: Interpreting, enforcing and changing the GNU GPL, as applied to combining Linux and ZFS

Posted Apr 13, 2016 10:36 UTC (Wed) by nye (guest, #51576) [Link]

>I think it's fair to ask Oracle to clean up the mess by simply putting out a statement saying that it is permissible to use ZFS under the GPL as well as the CDDL, and think it would be wise for Oracle to go along with this

Oracle ZFS is a proprietary product, and any open source ZFS is a direct competitor. There is essentially no chance that Oracle will choose to relicense ZFS under a GPL compatible license; the only circumstances in which I could imagine it happening are when they have a) given up on Solaris completely, and probably b) distributed ZFS with Linux and been sued over it. Even then, with Oracle's ZFS under a GPL compatible license, if you wanted that license to apply to the currently open source ZFS, you'd need to get agreement from all the contributors over the last 6(?) years and counting.