GPL and Linux loadable kernel modules [LWN.net]

GPL and Linux loadable kernel modules

Posted Aug 12, 2005 16:28 UTC (Fri) by giraffedata (subscriber, #1954)
In reply to: Interview: Eben Moglen by GreyWizard
Parent article: Interview: Eben Moglen

... end user links the module into the kernel, creating a work that can't be legally distributed
It's rather similar to the arguments Napster tried to use,

Actually, it's simpler than that. Although I've never seen how it fits into copyright law, US copyright law gives the author the right to control "preparation of derivative works." Even if you prepare it just for your own use, apparently. Stallman thought that was one of the greater threats against software freedom. I remember the argument being used against a person who was editing people's personal copies of Titanic video tapes to remove the R rated content (but I don't know if it succeeded).

But I'm surprised that Moglen believes everyone agrees that GPL doesn't permit linking non-GPL modules dynamically. From reading LWN, I believe it's a well established fact that there's disagreement on that.

Of course, we don't actually know Moglen believes that. We just know that he said he does. At least one other place in the interview, he uses rhetoric over clarity. He says he didn't say anything about SCO's reasons for leaving GNU alone, then makes it clear in the next sentence, using the word "because" instead of "reason," that he did.

(Log in to post comments)

GPL and Linux loadable kernel modules

Posted Aug 12, 2005 17:22 UTC (Fri) by GreyWizard (guest, #1026) [Link]

Actually, it's simpler than that. Although I've never seen how it fits into copyright law, US copyright law gives the author the right to control "preparation of derivative works."

Simpler? That seems more complex to me because going after end users would be a nightmare, even if you're right about the law. (Are you sure you're not confusing copyright law with contracts in end user license agreements? Could you cite a source?) Why is that simpler than accusing the vendor who provides the module of inducing infringement, especially given that many have succeeded in similar actions?

Maybe, but the FSF has consistently explained their opinion that dynamic linking has the same effect as static linking for derivative works. A court ruling would be necessary to really settle the point. I don't know about you, but Moglen has a law degree and much relevant practice so his opinion on this is worth something. His position makes intuitive sense too, at least to me. The difference between dynamic and static linking is purely technical and doesn't seem to be relevant to whether something is a derivative work.

What does seem to be relevant -- and this is the heart of the issue -- is what Linus Torvalds has pointed out. When code desinged and published independently of Linux is ported it is hard to determine whether this is a derivative work of the kernel, if only because Torvalds and other kernel copyright holders have said so. Will those copyright holders tolerate binary only modules that implement specifications that are independent of Linux and don't stray beyond the GPL-only boundaries established by the module system? They have so far but that doesn't mean they always will. Moglen seems to want a reliable answer one way or another.

GPL and Linux loadable kernel modules

Posted Aug 13, 2005 3:42 UTC (Sat) by giraffedata (subscriber, #1954) [Link]

I didn't read your post closely enough. Looking at it now, I see that you're not addressing the parent poster's claim that doing an insmod isn't a copyright violation. Instead, you're talking about the guy who supplied the LKM to insmod violating copyright. However, without establishing that doing the insmod is a copyright violation, I don't see how you could possibly argue that supplying the LKM violates copyright law by inducement.

In any case, there's nothing simple about making an inducement argument. This is fairly new law, and judge-made (that's why there was so much uncertainty about how Napster would turn out). By contrast, the derivative work law starts out in statute, and has been refined for about 100 years.

Check out the US Copyright statute, Title 17 of United States Code. (Just google "derivative work"). It lists the 6 exclusive rights that copyright law creates. #2 is "the right to prepare derivative works based on the copyrighted work." It is, of course still up in the air whether insmod qualifies as preparing a derivative work, but you can find plenty written on the idea that modifying a program for your own use qualifies. GPL enumerates preparing a derivative work as a specific right the licensee gets.

I agree that Eben Moglen and FSF believe and have always believed that insmodding is within those exclusive rights of the copyright holder. I only commented on what Moglen said about what everybody else believes.

By the way, Moglen's law degree doesn't mean squat here. A law degree from a US law school means you know the fundamental concepts of US law and have a broad overview of various areas of law -- all possibly as of 40 years ago. It doesn't mean you know the law of any particular jurisdiction or enough about any area of law to avoid embarrassing yourself in front of a judge. I know because I have one. BUT: I respect Eben Moglen's opinion because of his extensive study of copyright law and practice in the area. I have none (what I know about copyright law is mostly via LWN. I did learn some in law school, but have forgotten all that).

GPL and Linux loadable kernel modules

Posted Aug 13, 2005 20:09 UTC (Sat) by GreyWizard (guest, #1026) [Link]

Looking at it now, I see that you're not addressing the parent poster's claim that doing an insmod isn't a copyright violation.

Well, in fact I was though perhaps not clearly enough. My point was that attempting to push responsibility for binary modules from a vendor to end users, which is what we are doing when we reframe the problem in terms of insmod, is not likely to succeed. This is why I made the comparison to Napster and Grokster.

However, without establishing that doing the insmod is a copyright violation, I don't see how you could possibly argue that supplying the LKM violates copyright law by inducement.

If a binary kernel module is a derivative work of the kernel then distributing it is copyright infringement unless those who hold the kernel copyrights give permission, which they have not done by releasing under the GPL. Creating it in the first place is explicitly permitted by the GPL. So I guess I didn't read your post closely enough either: rather than calling your approach complex I ought to have said it was simply doomed. I also mispoke, perhaps still thinking of the comparison to Napster, when I mentioned inducing infringement. Distribution of derivate works is what I meant.

Perhaps now you understand why I think this is simpler than fretting about whether invoking insmod can be copyright infringement. Even without the GPL that hinges on strange questions about fair use and the way computers work -- not to mention the public relations problem suing potential customers represents. The distribution angle seems cut and dried by comparison. Both approaches require settling thorny questions about derived works which, not to put too fine a point on it, is the original topic here.

By the way, Moglen's law degree doesn't mean squat here. [...] BUT: I respect Eben Moglen's opinion because of his extensive study of copyright law and practice in the area.

Here is the text you are responding to: "I don't know about you, but Moglen has a law degree and much relevant practice so his opinion on this is worth something." The law degree is indeed important here, if only because without one such relevant practice would be illegal.

GPL and Linux loadable kernel modules

Posted Aug 13, 2005 20:32 UTC (Sat) by giraffedata (subscriber, #1954) [Link]

The law degree still isn't relevant because it isn't relevant to your argument whether the law practice of which you speak is legal.

There are cases of people practicing law illegally for long periods of time, and people generally admit that these people have, and have gained through the practice, expertise equivalent to what legal practitioners have.

Incidentally, there are still some states where there are ways to practice law legally without a degree. You always have to have a license (except in the case of some elected judgeships), but a degree isn't always a prerequisite to a license.

GPL and Linux loadable kernel modules

Posted Aug 15, 2005 14:25 UTC (Mon) by GreyWizard (guest, #1026) [Link]

The law degree still isn't relevant because it isn't relevant to yourargument whether the law practice of which you speak is legal.

I didn't know that one could get a license to pratice law without a law degree, so thank you for enlightening me on that point. Still, my error in conflating the two doesn't undermine the point I was making (since you start with "incidentally" I suppose you agree). Similarly, those who do practice law illegally probably have less experience on the whole than their licensed colleagues who are less likely to be jailed. At any rate claiming someone has credibility because they practice illegally probably isn't a useful argument.

I mentioned Moglen's law degree in the context of supporting his opinion on dynamic linking and derivative works. I did not say such support was absolute, that it should end the discussion or that people without law degrees cannot be right where he is wrong. I also accept that someone who practices law but works as, say, a divorce attorney or public defender might know less about copyright law than someone who reads LWN obsessively. Nevertheless it seems obvious that someone who spends time studying the law is in general more likely to have a correct opinion on a legal question than someone who doesn't, at least in the absence of more specific details. Thus a law degree lends legitimate credibility to opinions about the law, just as being any kind of trained professional lends credibility to opinions about that particular domain. I stand by my original statement.

GPL and Linux loadable kernel modules

Posted Aug 13, 2005 16:43 UTC (Sat) by Duncan (guest, #6647) [Link]

> But I'm surprised that Moglen believes
> everyone agrees that GPL doesn't permit
> linking non-GPL modules dynamically.
> From reading LWN, I believe it's a well
> established fact that there's
> disagreement on that.

First, not to be niggling, but as one who claims to have a law degree
(which I'm not disputing), you should know how much difference similar but
not identical words can make. Moglen did /not/ say he believes "everyone"
agrees... What he DID say is "I think we would all agree"... (That's a
direct select/paste.)

The significance of "we would all agree" vs. "everyone agrees" should be
quite apparent. One, "everyone", without a specific exception appended,
tends to paint one into a corner, because it's /very/ easy to disprove,
with even a /single/ exception. "We" is a plural pronoun, and even
modified by "all", remains vague enough to allow exceptions. In
particular, "we" implies a group of which he considers himself a part, but
the group isn't specifically defined within the interview. Within
context, particularly inclusive of the second factor below, I took the
"we ... all" to mean "informed involved parties." Specifically, that
would mean "kernel maintainers and other significant contributors who have
taken the time to become informed on the subject, together with legal
council." That specifically excludes "everyone", altho because "we"
wasn't specifically defined, that remains personal opinion, the way /I/
took it.

Still considering the phrase "we would all agree", there's the term
"would". Within context, that lends support to the above "informed"
interpretation. Specifically, "would" implies some qualification or
condition, that "we all agree" /if/ said qualification "would" be met.
Within context, the implication is that "we would all agree" ... /if/ we
all took the time to study the issues...

Together, then, the two parts of the phrase he chose do /not/ necessarily
imply the all-inclusive "everyone" that was your interpretation. Rather,
the phrasing used indicates a somewhat more selective group, and only then
conditionally, with the implied condition being that of having taken the
time to be informed on the subject. Being rather less all-inclusive, the
statement he made (as quoted), as opposed to the statement you attributed
to him, is significantly easier to defend, and to understand.

Second, and similar to you, from my reading of LWN (and other community
sources such as NewsForge), the "well established" disagreement appears to
me to be specifically with the conditions applied to the Linux
kernel, /not/ to the GPL in general.

Again, the distinction can be subtle, but it's quite important,
particularly so in understanding the point Moglen was (IMV) trying to
make. The Linux kernel is an exception in GPL licensed software, due to
the fact that its license, as contained within the COPYING file, contains
a preamble to the GPL. Go ahead and look it up. It's worth reading.

The point Moglen was making was this: The GPL, as applied to the kernel,
already has specific exception clauses (or clarifications, as the case may
be argued) applied, so it's /not/ /just/ the terms of the GPL that apply,
it's the GPL already modified. The "informed" as implied by the "we would
all agree" as discussed above, will already understand this Linux kernel
exception, and read his comments with it in mind.

Further to the point, there was honest debate as to whether userland was
covered by the code sharing requirements of the GPL, particularly where
such userland made use of kernel headers. Since glibc makes use of kernel
headers, and most of userland in turn links against it, one could argue
that nearly the entire Linux userland was subject to the terms of the GPL.
That didn't fit Linus' idea of the conditions he wanted for the kernel at
all, so he put an end to the debate with the GPL preamble in the COPYING
file, specifically allowing proprietary source userland applications.

It is within the context of this history that Moglen makes his comments to
the effect that were the license pure GPL /only/, without the preamble and
other to this day not specifically stated exceptions, he thinks "we would
all agree" that kernel linking, static or dynamic, would violate the GPL.
His argument is that such other exceptions need to be stated specifically,
allowed or disallowed with a clear delineation, as has already been made
with the preamble and with the GPL-only function declarations.

It is within that context that I read his arguments, and believe that he's
right, were the GPL /only/ to be applied, there'd be little controversy,
all closed kernel modules would be a violation of the license, period.
The reason there's controversy at all, is due to the very reasonable
exceptions already made, along with hints, both in Linus' remarks, and in
the written form of the specifically GPL-only kernel symbols, that other
not-specifically-delineated exceptions exist as well. Because these
exceptions apply specifically to the kernel, not to the GPL or other GPL
licensed software, the controversy isn't on the GPL itself, and I believe
his statement that "we would all agree", within the context outlined, is a
fairly safe statement to make.

All in my opinion as specifically NOT a lawyer, and NOT a kernel level
coder, either, just a (hopefully) astute consumer of legal opinion on the
topic at hand.

Duncan