Index Home About Blog
From: Linus Torvalds <>
Newsgroups: fa.linux.kernel
Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3
Date: Fri, 15 Jun 2007 15:46:59 UTC
Message-ID: <fa./>

On Fri, 15 Jun 2007, Carlo Wood wrote:

> On Fri, Jun 15, 2007 at 06:33:51AM -0400, Daniel Hazelton wrote:
> > Incorrect. Read section 9 of the GPLv2. It's pretty clear that the "any later
> > version" clause is optional. Whats more is that since the modern linux kernel
> > *IS* a "composite work" composed of Linus' original code with changes
> > contributed by other people - Linus retains copyright to the work as a whole.
> Huh - surely not to files added to the kernel that were written by
> others from scratch!

Actually, yes. Even to those - when they are part of "the whole".

I'm sorry, but I've learnt more about copyright law, and talked to more
lawyers about licensing that probably most of the rest of the people
involved in this discussion have *combined*.

And yes, at least under US copyright law, and at least if you see Linux as
a "collective work" (which is arguably the most straightforward reading og
copyright law, but perhaps not the only one) I am actually the sole owner
of copyright in the *collective* work of the Linux kernel.

The way "collective works" work, there are two separate copyrights: there
is the copyright in the "separate contribution", which is vests ininitally
in the author of that contribution (unless he signs over his copyrights,
often by virtue of working for somebody else).

And then there is the copyright in the "collective work", which would be

Of course, owning coyright in the "collective work" doesn't actually give
me complete control anyway. I cannot relicense things in ways that go
against the rules of the individual works. But in a very real sense, yes,
I actually do own a certain (*limited*) copyright over even the parts that
have not been explicitly signed over to me.

And yes, there are other potential ways to describe Linux, and in the end,
it doesn't really matter. Because the way the GPLv2 works, it makes it
clear that as long as a piece is a part of the whole, it has to be
licensed under the GPLv2 and nothing else.

And btw, just to make you feel safe - I cannot do anything about that,
even if I *do* own the copyright in the collective, because of the
limitations on what that colletive work copyright implies (it says that I
have the right to reproduce and distribute, but I don't have the right to
*modify* except as given to me by the original author!)

So don't worry. I *technically* have certain special rights, but I
practically speaking gave up most all of those rights by accepting code
from others under the GPLv2 - since in order to do that, I had to agree to
be bound by the GPLv2 license myself.

> This is totally new to me - if this is true - I'd really like to be sure!

It would be generally held to be true at least in the US, but it doesn't
really matter.

> I always thought that it would be necessary to get signatures of each
> and every contributor before you can change a license of a file.

You are mostly correct. The "mostly" comes because I would not say "every
contributor", but would clarify it by saying "every copyright holder". The
difference? Not all contributions are necessarily copyrightable. If you
send in trivial one-liners, we will credit you for them, but that does not
automatically mean that you necessarily own copyright in something.

But yes, somebody who wrote an original file (that has some artistic
expression, and isn't just a list of PCI ID's, for example) will be the
copyright owner in that file. Some *very* few people have actually sent me
paperwork to transfer the ownership of copyrights, but they seem to have
done that because they were just used to doing it with the FSF, and I
actually don't care.


From: Linus Torvalds <>
Newsgroups: fa.linux.kernel
Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3
Date: Fri, 15 Jun 2007 18:00:52 UTC
Message-ID: <>

On Fri, 15 Jun 2007, Adrian Bunk wrote:
> US law is only relevant for < 5% of all people.
> How valid would any action based on US copyright law be in other parts
> of the world?

I tried to explain that in the case of the Linux kernel, we really don't
care, since in the end, what matters is the GPLv2, and I have bound myself
to the terms of that license *regardless* of any US law.

So yes, US law is only relevant for < 5% of all people, but in the
specific case of the kernel, even that US law isn't _really_ all that
relevant at all, not *even* to those 5% ;)

So I *really* hope you took my explanation of why I actually have more
rights than others as a nitpicking "legal detail", not as a "I own your
very SOUL, bow down before me!" kind of thing.

But to answer your question by _another_ nitpicking answer, as the
original author, I probably do have some special legal status even in
Europe, and probably in other places too. The fact that others *extended*
on my original work doesn't take away the special place of original
authorship, even if the extended version has a totally different form (ie,
a movie based on a book ends up still having the original author of the
book holding special rights - and in fact those rights are in some cases
much *stronger* in Europe than they are in the US).

For example, Europe recognizes "moral rights" in original authorship, in
ways that it is much harder to enforce (if at all) in the US.

But as mentioned, since I myself has bound myself to the GPLv2, that
really is a pretty damn theoretical argument. When it comes to the kernel,
I'm "Primus inter pares", if you wish.

("First among equals", for the non-latin-speaking world ;)

> Does this include GPLv2'ed code not intended to be used in the Linux
> kernel submitted by people other than the copyright holder for inclusion
> in the Linux kernel?

In the US sense of "within the compilation", probably yes. Ie that right
is tied to *linux* as a compilation. In the European sense of "moral
rights", no - that right is very much tied to original authorship (so I,
as original author of Linux, have some rights with respect to Linux, but
on the other hand, they, as the original authors of some specific code,
have some specific rights wrt that code!)

> If yes, the FSF has exactly the same rights if taking a GPLv2 driver
> from the Linux kernel and including it in GNU Hurd.

The FSF does indeed have special rights wrt Hurd, regardless of where the
code in question has come from (as long as it came there *legally*, of
course!). So when it comes to Hurd, they have the copyright in the
compilation of that (and Red Hat as their copyright in their RHEL
distribution - of which the kernel is just a small part!).

It's a confusing set of nesting dolls of different levels of copyrights,
and the GPLv2 makes it all pretty much irrelevant, by being the thing that
cuts through all the other licensing issues!


From: Linus Torvalds <>
Newsgroups: fa.linux.kernel
Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3
Date: Fri, 15 Jun 2007 18:17:24 UTC
Message-ID: <>

On Fri, 15 Jun 2007, Carlo Wood wrote:
> The point is: can you, or can't you (legally) relicense the whole kernel
> tree under the GPLv3 (or GPLv2+GPLv3)?

No. My special rights do not actually give me those kinds of powers,
exactly because I'm bound by my _other_ agreement (namely the GPLv2) to
follow the license of the code that other people have sent me.

> At first I thought that you cannot, because too many (significant) contributors
> have been involved (and you will never get signatures from them all).
> Then someone surprised me by claiming that the original author had
> copyright for everything - even files added by others.

Both are true facts, but the "copyright for everything" is a *separate*
kind of copyright, which does not include the right to relicense. It's
literally the "copyright in the collective".

For examples of the US rules, see USC 17.2.201(c) ("Ownership of
copyright" and " Contributions to Collective Works"), which spells out
some limited special rights that I have (namely the right to reproduce and

Of course, US law being what it is, the USC is just part of the picture.
US law is the strange kind of British law, where "case law" is in many
ways more important than the written-down rules like the USC. So caveat

So I have limited special rights in the collective, but those rights are
actually in almost every way *more* limited than the rights that the GPLv2
gives to me (the "almost every way" is because quite frankly, I'm not
entirely sure about certain special cases. In particular, if somebody
tried to _revoke_ the rights to their code under the GPLv2, I suspect that
my rights in the collective would protect me from that and allow me to
still distribute the code in question, since _those_ rights cannot be
revoked, and they are _mine_).

And btw: the above paragraph is *way* more legalistic detail than I am at
all ready to state as "fact". It depends on too many things, and is
largely speculative in nature.

But one thing is pretty clear and nonspeculative: *nobody* has the right
to upgrade the kernel to GPLv3. Not me, not you, not anybody. Not without
clearing it with every single person whose copyright is involved and who
didn't already give that permission.

So only in the case of some really obscure and unclear situations, I _may_
have more rights than some other people, but trust me, but that is damn
murky, and you'd better have a good lawyer state it, not just a programmer
who has talked to too many lawyers..


Index Home About Blog