"I was impressed in the sense that it was a hell of a lot better than the disaster that were the earlier drafts," Linus Torvalds explained in reply to a comment suggesting that he was impressed with the final draft of the GPLv3. He went on to add, "I still think GPLv2 is simply the better license." The discussion began with a suggestion that the Linux kernel be dual-licensed GPLv2 and GPLv3. Linus noted, "I consider dual-licensing unlikely (and technically quite hard), but at least _possible_ in theory. I have yet to see any actual *reasons* for licensing under the GPLv3, though. All I've heard are shrill voices about 'tivoization' (which I expressly think is ok) and panicked worries about Novell-MS (which seems way overblown, and quite frankly, the argument seems to not so much be about the Novell deal, as about an excuse to push the GPLv3)." In a followup email, Linus added:
"Btw, if Sun really _is_ going to release OpenSolaris under GPLv3, that _may_ be a good reason. I don't think the GPLv3 is as good a license as v2, but on the other hand, I'm pragmatic, and if we can avoid having two kernels with two different licenses and the friction that causes, I at least see the _reason_ for GPLv3. As it is, I don't really see a reason at all."
From: Tarkan Erimer [email blocked] To: linux-kernel Subject: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sat, 09 Jun 2007 08:46:46 +0300 Hi, As we know the forthcoming GPL V3 will be not compatible with the GPL V2 and Linux Kernel is GPL V2 only. So, another point is, which is previously mentioned by Linus and others, that if it is decided to upgrade the Linux Kernel's License to GPL V3, it is needed the permission of all the maintainers permission who contributed to the Linux Kernel and there are a lot of lost or dead maintainers. Which makes it impossible to get all the maintainers' permission. But; if the Linux kernel should Dual-Licensed (GPL V2 and GPL V3), it will allow us the both worlds' fruits like code exchanging from other Open Source Projects (OpenSolaris etc.) that is compatible with GPL V3 and not with GPL V2 and of course the opposite is applicable,too. So;at this situation, what is possibility to make the Linux Kernel Dual-Licensed as I mentioned above and what is your opinions and suggestions about this idea ? Regards, Tarkan Erimer
From: Neil Brown [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sat, 9 Jun 2007 15:57:55 +1000 On Saturday June 9, [email blocked] wrote: > Hi, > > As we know the forthcoming GPL V3 will be not compatible with the GPL V2 > and Linux Kernel is GPL V2 only. > So, another point is, which is previously mentioned by Linus and others, > that if it is decided to upgrade the Linux Kernel's License to GPL V3, > it is needed the permission of all the maintainers permission who > contributed to the Linux Kernel and there are a lot of lost or dead > maintainers. Which makes it impossible to get all the maintainers' > permission. You don't need the permission of maintainers. You need the permission of copyright owners. The two groups overlap, but are not the same. Dead people cannot own anything, even copyright. Their estate probably can. I don't think it is theoretically impossible to get everyone's permission, though it may be quite close to practically impossible. > But; if the Linux kernel should Dual-Licensed (GPL V2 and GPL V3), it > will allow us the both worlds' fruits like code exchanging from other > Open Source Projects (OpenSolaris etc.) that is compatible with GPL V3 > and not with GPL V2 and of course the opposite is applicable,too. > > So;at this situation, what is possibility to make the Linux Kernel > Dual-Licensed as I mentioned above and what is your opinions and > suggestions about this idea ? Dual licensing is no easier. It means it is licensed to be used under either license. You already have permission to use it under GPLv2. So to get a dual license, you precisely need to get access under GPLv3 i.e. to convince copyright owners to make that license grant. A thing that we have already agreed is at least "hard". NeilBrown
From: Jan-Benedict Glaw [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sat, 9 Jun 2007 09:12:31 +0200 On Sat, 2007-06-09 15:57:55 +1000, Neil Brown [email blocked] wrote: > On Saturday June 9, [email blocked] wrote: > > As we know the forthcoming GPL V3 will be not compatible with the GPL V2 > > and Linux Kernel is GPL V2 only. > > So, another point is, which is previously mentioned by Linus and others, > > that if it is decided to upgrade the Linux Kernel's License to GPL V3, > > it is needed the permission of all the maintainers permission who > > contributed to the Linux Kernel and there are a lot of lost or dead > > maintainers. Which makes it impossible to get all the maintainers' > > permission. > > You don't need the permission of maintainers. You need the permission > of copyright owners. The two groups overlap, but are not the same. > Dead people cannot own anything, even copyright. Their estate > probably can. I don't think it is theoretically impossible to get > everyone's permission, though it may be quite close to practically > impossible. And the next question is: How much copyright does a copyright owner own? For example, think of drivers written by one person, but a small number of lines changed here and there by others to adopt the code to new APIs. Ask them all, I think? MfG, JBG -- Jan-Benedict Glaw [email blocked] +49-172-7608481 Signature of: God put me on earth to accomplish a certain number of the second : things. Right now I am so far behind I will never die.
From: Tarkan Erimer [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sun, 10 Jun 2007 11:43:28 +0300 Jan-Benedict Glaw wrote: > On Sat, 2007-06-09 15:57:55 +1000, Neil Brown [email blocked] wrote: > >> On Saturday June 9, [email blocked] wrote: >> >> You don't need the permission of maintainers. You need the permission >> of copyright owners. The two groups overlap, but are not the same. >> Dead people cannot own anything, even copyright. Their estate >> probably can. I don't think it is theoretically impossible to get >> everyone's permission, though it may be quite close to practically >> impossible. >> > > And the next question is: How much copyright does a copyright owner > own? For example, think of drivers written by one person, but a small > number of lines changed here and there by others to adopt the code to > new APIs. Ask them all, I think? > > MfG, JBG > > And maybe another questions should be : How long a copyright owner can hold the copyright, if died or lost for sometime ? if died, the copyright still should be valid or not ? If lost, what the law orders at this point for copyright holding ?
From: Tarkan Erimer [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sun, 10 Jun 2007 13:00:28 +0300 [email blocked] wrote: > On Sun, 10 Jun 2007, Tarkan Erimer wrote: > >> Date: Sun, 10 Jun 2007 11:43:28 +0300 >> From: Tarkan Erimer [email blocked] >> Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 >> >> And maybe another questions should be : How long a copyright owner >> can hold the copyright, if died or lost for sometime ? if died, the >> copyright still should be valid or not ? If lost, what the law orders >> at this point for copyright holding ? > > I believe that in the US it's life + 90 years. > > David Lang Hmm... Really,it is damn too much time to wait! It's really better idea to replace the code of this person as said before instead of waiting such 90+ years!
From: [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sun, 10 Jun 2007 03:03:25 -0700 (PDT) On Sun, 10 Jun 2007, Tarkan Erimer wrote: >> I believe that in the US it's life + 90 years. >> >> David Lang > Hmm... Really,it is damn too much time to wait! It's really better idea to > replace the code of this person as said before instead of waiting such 90+ > years! exactly, however as others are pointing out, there are a lot of active developers who do not agree with some of the key points of the GPLv3 (including Linus), so until you convince them that the GPLv3 is better it really doesn't matter how hard it is to deal with the people who you can't contact. David Lang
From: "debian developer" [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sun, 10 Jun 2007 16:25:55 +0530 On 6/10/07, David Lang wrote: > On Sun, 10 Jun 2007, Tarkan Erimer wrote: > > >> > And maybe another questions should be : How long a copyright owner can > >> > hold the copyright, if died or lost for sometime ? if died, the > >> > copyright still should be valid or not ? If lost, what the law orders at > >> > this point for copyright holding ? > >> > >> I believe that in the US it's life + 90 years. > >> > >> David Lang > > Hmm... Really,it is damn too much time to wait! It's really better idea to > > replace the code of this person as said before instead of waiting such 90+ > > years! > > exactly, however as others are pointing out, there are a lot of active > developers who do not agree with some of the key points of the GPLv3 > (including Linus), so until you convince them that the GPLv3 is better it Last heard, Linus was quite impressed with the toned down version of the final draft of GPLv3. I think Linus, and other major developers should make their stand on this issue clear so that the kernel community can discuss the future steps.
From: Tarkan Erimer [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sun, 10 Jun 2007 17:21:53 +0300 debian developer wrote: > On 6/10/07, David Lang wrote: >> On Sun, 10 Jun 2007, Tarkan Erimer wrote: >> >> exactly, however as others are pointing out, there are a lot of active >> developers who do not agree with some of the key points of the GPLv3 >> (including Linus), so until you convince them that the GPLv3 is >> better it > > Last heard, Linus was quite impressed with the toned down version of > the final draft of GPLv3. I think Linus, and other major developers > should make their stand on this issue clear so that the kernel > community can discuss the future steps. Yep, the GPLv3 probably will release around July time. So;luckily, we had very little time to see the final decision about it :-) I hope we should upgrade to GPLv3 and Sun should "Dual License" the OpenSolaris via GPLv3 (or at least,GPLv3 should be CDDL compatible.). So,we should have more fruits (like ZFS,DTrace etc.) ;-)
From: Linus Torvalds [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sun, 10 Jun 2007 10:29:04 -0700 (PDT) On Sun, 10 Jun 2007, Tarkan Erimer wrote: > > > > Last heard, Linus was quite impressed with the toned down version of > > the final draft of GPLv3. I was impressed in the sense that it was a hell of a lot better than the disaster that were the earlier drafts. I still think GPLv2 is simply the better license. I consider dual-licensing unlikely (and technically quite hard), but at least _possible_ in theory. I have yet to see any actual *reasons* for licensing under the GPLv3, though. All I've heard are shrill voices about "tivoization" (which I expressly think is ok) and panicked worries about Novell-MS (which seems way overblown, and quite frankly, the argument seems to not so much be about the Novell deal, as about an excuse to push the GPLv3). Linus
From: Linus Torvalds [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sun, 10 Jun 2007 10:33:54 -0700 (PDT) On Sun, 10 Jun 2007, Linus Torvalds wrote: > > I have yet to see any actual *reasons* for licensing under the GPLv3, > though. Btw, if Sun really _is_ going to release OpenSolaris under GPLv3, that _may_ be a good reason. I don't think the GPLv3 is as good a license as v2, but on the other hand, I'm pragmatic, and if we can avoid having two kernels with two different licenses and the friction that causes, I at least see the _reason_ for GPLv3. As it is, I don't really see a reason at all. I personally doubt it will happen, but hey, I didn't really expect them to open-source Java either(*), so it's not like I'm infallible in my predictions. Linus (*) And I've been pushing for that since before they even released it - I walked out on Bill Joy at a private event where they discussed their horrible previous Java license.
From: Tarkan Erimer [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Mon, 11 Jun 2007 11:38:42 +0300 Linus Torvalds wrote: > On Sun, 10 Jun 2007, Linus Torvalds wrote: > >> I have yet to see any actual *reasons* for licensing under the GPLv3, >> though. >> > > Btw, if Sun really _is_ going to release OpenSolaris under GPLv3, that > _may_ be a good reason. I don't think the GPLv3 is as good a license as > v2, but on the other hand, I'm pragmatic, and if we can avoid having two > kernels with two different licenses and the friction that causes, I at > least see the _reason_ for GPLv3. As it is, I don't really see a reason at > all. > > I personally doubt it will happen, but hey, I didn't really expect them to > open-source Java either(*), so it's not like I'm infallible in my > predictions. > > Linus > > (*) And I've been pushing for that since before they even released it - I > walked out on Bill Joy at a private event where they discussed their > horrible previous Java license. > Thanks for making things more clear :-) Some really strong indications that Sun is very willing to,at least, "Dual-License" the OpenSolaris with GPLv3. I think; in a very short time; we will see when the GPLv3 finalized and released. Regards, Tarkan
From: Ingo Molnar [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Mon, 11 Jun 2007 11:03:48 +0200 * Tarkan Erimer [email blocked] wrote: > > (*) And I've been pushing for that since before they even released > > it - I walked out on Bill Joy at a private event where they > > discussed their horrible previous Java license. > > Thanks for making things more clear :-) Some really strong indications > that Sun is very willing to,at least, "Dual-License" the OpenSolaris > with GPLv3. I think; in a very short time; we will see when the GPLv3 > finalized and released. that would certainly be a good and productive move from them. Note the issue that others have pointed out to you: OpenSolaris is probably more interested in picking up code from Linux than the other way around! :-) You mentioned "dtrace" and "ZFS". Firstly, Linux already has a "dtrace" equivalent. Secondly, ZFS might be interesting in theory, although our prior experience of having compatibly-licensed filesystems ported over to Linux has been pretty negative: XFS ended up being an integration nightmare - and that doesnt have to do anything with the qualities of XFS (it's one of the cleanest Linux filesystems, if not the cleanest), the problem is that components within a kernel are very tightly integrated and rarely does it make sense to port over more than just drivers or maybe libraries. And that's i guess what OpenSolaris lacks and which i suspect it is mostly interested in: lots of nice Linux drivers ;-) XFS, the largest Linux filesystem is 100K lines of code - and ZFS (i've never seen it) is very likely smaller than that. Linux drivers on the other hand, as of today, are _3.7 million_ lines of code and enable Linux to run on 99% of the hardware that is produced today. Guess which one has the larger strategic significance? ;-) Ingo
From: Alan Cox [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sun, 10 Jun 2007 18:47:36 +0100 Windsor, Berkshire, SL4 1TE, Y Deyrnas Gyfunol. Cofrestrwyd yng Nghymru a Lloegr o'r rhif cofrestru 3798903 > licensing under the GPLv3, though. All I've heard are shrill voices about > "tivoization" (which I expressly think is ok) and panicked worries about GPLv2 probably forbids Tivoisation anyway. Which is good IMHO even if not yours 8)
From: "debian developer" [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Mon, 11 Jun 2007 01:02:42 +0530 On 6/10/07, Alan Cox [email blocked] wrote: > > licensing under the GPLv3, though. All I've heard are shrill voices about > > "tivoization" (which I expressly think is ok) and panicked worries about > > GPLv2 probably forbids Tivoisation anyway. Which is good IMHO even if not ^^^^^^^^ Now that is a bit waving in the air. GPLv2 forbids Tivoisation theoretically but practically it didnt stop them doing it practically. I agree with Linus that software licenses should have their influence only on the software part and leave the freedom of the hardware on which the software runs to the hardware manufacturers. But was it the goal of GPLv2?? And what does Andrew Morton think of all this? I really want to know his opinions....
From: Andrew Morton [email blocked] Subject: Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 Date: Sun, 10 Jun 2007 13:02:18 -0700 On Mon, 11 Jun 2007 01:02:42 +0530 "debian developer" [email blocked] wrote: > And what does Andrew Morton think of all this? I really want to know > his opinions.... I have yet to see Linus make a statement on these matters with which I didn't agree.
So what happens when
So what happens if I as the copyright owner of a module which people use in the kernel, decide that I like GPL3. If I make my code GPL3 does that mean that existing versions within the kernel will automatically get updated with my next commit, or with the commit be rejected because of license issues?
You cant. GPL2 excludes any
You cant. GPL2 excludes any other version.
GPL3 cant be included and it is a break of license terms if your modules will require GPL3.
He can do whatever he wants
He can do whatever he wants as the copyright holder. It is a GPLv2 kernel which would be undistributable with a GPLv3 module. This is only a problem because the kernel license does not include the "... or any subsequent version" language which would have allowed it to be distributed with GPLv3 modules.
Re: So what happens when
In that case, the new code is not part of the GPLv2 kernel, but the existing GPLv2 code still is; you can't withdraw software once it is under GPLv2, you can only cease to maintain it. As a developer, you can walk away with your new code, and someone else can pick up maintenance of the existing code.
You can't revoke copies
If you release something under GPL 2, and someone makes a copy of it and stays within the requirements of GPL 2, there's nothing you can do to take that from them.
You, as a sole copyright holder, can issue your creative works under whatever license you like. So, you release one version under GPL 2 (w/out the "or later version" clause). Later you release a GPL 3 version. Guess what? Each is an independent release, and the licensing terms associated with them has nothing to do with the other release.
So, in your example... you write a module and license it under GPL 2. A copy gets incorporated in the kernel. Later, since you're the sole copyright owner in your driver, you decide to move to GPL 3. Your previous versions are still out there as GPL 2. It's only your new changes that get covered by GPL 3.
How does a dual-licensed
How does a dual-licensed kernel avoid tivoization? If I want to tivo-ize a kernel, then I just use it under GPLv2 as before.
Require all kernel must be "GPL2 or later" by a certain date.
Dual licensing in GPL2 and GPL3 won't ever happen unless there is a requirement to do so by a specific data. If the changeover is to ever take place there is a need to require all code submissions to be "GPL2 or later" by a certain date, which means GPL2 will be used for now, but the Linux kernel can be shifted over to GPL3 if required at some stage in the future.
Bearing in mind that the GPL licenses need to be changed from time to time, I think that perhaps it would be a good idea to change GPL so that it can be linked to code released under a later version of GPL even if the individual piece of code itself is only under the current version on GPL.
the cleanest path
The cleanest path to a GPL v3 Linux would be a rewrite, be it just the dubious portions, or even better a full one.
This will have several side-effects:
* many actually hidden bugs would be fixed
* refactorization can lead to leaner and cleaner code in some places
* the kernel can start from a clean attribution state
It, of course, could lead to new bugs. But then this would be an opportunity to write a good test harness for many kernel parts.
It would be expensive, of course. Maybe a couple of years or more. But in the end we will have a very good reason for a Linux 3.0 version.
And call it Linux Vista ?
And call it Linux Vista ?
A more realistic (slower) path
Since the demand for the GPLv3 isn't that high, I'd guess a more realistic path would be:
1) Owners of modules who like the GPLv3 would dual-license any new additions GPLv2 or GPLv3 and relicense any previous code the wrote or people agree to the dual-licensing wrote.
2) For each of these licenses, GPLv2-only code that could not be relicenses would be isolated in separate files. If no such GPLv2-only code exists, Go to step 4.
3) Eventually, at least some of that code may be written, and if it is, it would be dual-licensed as GPLv2 or GPLv3. Go to step 2.
4) Once all the GPLv2-only code is purged from the module, the module would be declared GPLv2 or GPLv3.
5) Once all modules are GPLv2 or GPLv3, the Linux kernel would be GPLv2 or GPLv3.
I believe Linus stated that there's almost no code in the kernel that has survived 5 years without rewriting, so given this, it should be possible to relicense the Linux kernel painlessly (with little effort or friction) within 5 years.
You can't do it piecemeal.
You can't do it piecemeal. Linux is currently GPLv2, de facto. If you just rewrite one module that uses internal interfaces, the new module is GPLv2 because it's a derived work of the kernel and must follow the kernel's license. So you're back where you started. In effect, all the GPLv2 components must be rewritten and relicensed simultaneously.
Of course you can
Of course you can do it module at a time. That's how Mozilla was relicensed to be GPL compatible.
The combined work would be GPLv2, *but* the individual modules would be dual licensed. There's BSD code in the kernel that's still under the BSD license mostly because the authors want their code to be shared with FreeBSD. That's not an issue since BSD licensed code is GPL compatible (i.e. it can be relicensed to GPL on demand).
So while 99.99% GPLv2+v3 code plus .01% GPLv2 only code would give you GPLv2 only code, once you remove or replace the .01% GPLv2, the whole code base will be GPLv2+v3.
The difference between
The difference between Mozilla and Linux is that the Mozilla developers did not publicly state that they consider any module using internal interfaces as a derived work, where the Linux developers did. How can you license a derived work of a GPLv2 product under a GPLv3 license? You can't. The Linux developers will have to drop their derived work stance in order to make this work, which calls into question the whole MODULE_LICENSE() scheme.
I don't think you
I don't think you understand.
The copyright holder gets to allow the copying of his/her code in as many independent ways as they want to. If they say this code can be copied under either GPL2 or GPL3 terms, then it can be copied under either terms. Obviously if it is included in a project that is only compatible with GPL2, it is currently being copied around with that code under the terms of the GPL2. This does not stop it, by itself, from being copied under the GPL3 terms at any point. I could freely rip it off into my independent GPL3 only project at any time.
Therefore, if all the code in the entire kernel eventually states it can be copied under either the GPL2 or GPL3, it suddenly becomes possible to copy the work as a whole under the terms of the GPL3. This achieves a transition. The addition of GPL3 only components would then complete the transition, as it would no longer be possible to copy the work as a whole under the terms of the GPL2.
No, that's not how it works.
No, that's not how it works. Any component designed to replace a component of the existing GPLv2 kernel is considered a derived work of the kernel. It is thus subject to GPLv2 and CANNOT be relicensed as GPLv3 as it is a derived work.
The kernel team is going to have to drop their stance on modules using internal interfaces being derived works in order for this scheme to work. Of course in the process we would lose the protections against proprietary modules hijacking kernel internals and giving Linux a bad reputation with their bugs.
Does any *license* say internal interface use == derived work?
"Any component designed to replace a component of the existing GPLv2 kernel is considered a derived work of the kernel."
What is the legal basis for this? AFAIK, "derived work" means that it contains code copyrighted by someone else.
Consider that the SCOundrels have been trying hard to argue that "interfaces", "methods", "concepts" and so forth make Linux a derivative work of Unix. They are going to fail, because what they really need to demonstrate is that Linux contains unauthorized Unix code, which it almost certainly does not.
The kernel devs can "take their stance" all they want, but I don't think they can legitimately claim that all code *added to* Linux is therefore *derived from* Linux. It might have to be licensed under the GPL2 to be distributed with the GPL2 kernel, but the author of the new code most definitely has the right to release it under other licenses as well, including the GPL3. The author just would not have the right to release any code *taken from* GPL2 kernel code and release it under another license, unless he/she gets permission from *that* author.
Your argument is just like SCO claiming that JFS became Unix code after IBM contributed it to AIX, and therefore IBM lost the right to contribute it to Linux.
No, if it's a DERIVED work,
No, if it's a DERIVED work, the "author of the new code" most definitely DOES NOT have the right to release it under other licenses as well.
Educate yourself:
http://kororaa.org/static.php?page=gpl
But it isn't "derived"
Sure, a derived work couldn't be relicensed under GPLv3 without getting the upstream author to agree. However, new code that is not derived from previous GPLv2 code can be released under whatever license or licenses the author wishes. My argument is that nothing in the GPLv2 would cause a new kernel module to be considered a derived work.
(see below)
This isn't about whether a
This isn't about whether a "new kernel module" in the sense of a new driver would be considered a derived work. This is about designing a drop-in replacement for existing kernel code. Since there is no published API boundary, and the code would do nothing on any system except Linux, there is a strong argument -- supported by the kernel team's statements -- that such a work would be considered derived.
If, OTOH, it is *not* a derived work, then that certainly means that NVIDIA and the likes are free to muck with kernel internals all they want, with the foreseen consequences in terms of frustrated developers and a bad reputation for Linux.
one other thing
The link you provided just gives the opinions of Linus and the other kernel devs. It is not backed up by the wording of the license itself, nor of any legal decisions, nor of any legal opinions from actual lawyers (not that I'm one, which I am not).
It just seems to me that considering all modules to be "derived works" is not backed up by the wording of the actual license, is certainly counter-intuitive, and might be very shaky to try to enforce in court. I strongly doubt the courts would care whether something was "primarily developed for" the linux kernel, as opposed to being ported from some other system (like JFS). The relevant law is *copyright*, remember, and copyright has to do with copying of text, not the purpose for which the text was created. If a new module does not contain old code, how can it be a *copyright* violation? And if it isn't a *copyright* violation, what other legal basis is there for saying the author can't release it under whatever license the author sees fit? I don't think the "stance" of Linus and the other kernel devs is the deciding factor here.
If you copy something and
If you copy something and reword it, but the structure and idea of the work bears substantial similarity to the source, it is a derived work. Same for rewriting a replacement core kernel module just for the sake of relicensing.
No, if it's a DERIVED work,
No, if it's a DERIVED work, the "author of the new code" most definitely DOES NOT have the right to release it under other licenses as well.
Educate yourself:
http://kororaa.org/static.php?page=gpl
JFS was not developed
JFS was not developed primarily as a component of the Linux kernel. Therefore it is not a derived work. A replacement GPLv3 kernel module is developed primarily as a component of the Linux kernel. Therefore it is a derived work of the Linux kernel and must be placed under GPLv2, making it impossible to release as GPLv3.
Again, what in the license says such a module would be DERIVED?
"A replacement GPLv3 kernel module is developed primarily as a component of the Linux kernel. Therefore it is a derived work of the Linux kernel..."
Again, is there any legal basis for your statement? I have read the GPLv2 in entirety, and I haven't found anything about "developed primarily as a component" as a criterion for saying that a kernel module is a derived work.
Below is the exact text of what GPLv2 says constitutes a derivative work:
"...any derivative work under copyright law:
that is to say, a work containing the Program or a portion of it,
either verbatim or with modifications and/or translated into another
language."
Note that the license specifically says that a "derivate work" is one considered derivative under copyright law. This means that use of interfaces, concepts, methods, etc. do not make something a derivative work. The license doesn't say anything about whether the code in question was developed primarily to work with GPL code.
So, please explain just what would make a new kernel module a derivative work, if it is written from scratch without using any existing kernel code?
Why are you referring to the
Why are you referring to the GPL as an authority on what constitutes a derivative work? The GPL cannot determine that. Only the copyright statutes and case law can. We don't have case law, we only have a gentlemen's agreement from the kernel leads, so either that agreement must be respected or it must be renegotiated.
http://www.wasabisystems.com/Technology/GPL/LKMs_as_%22Derivative_Works%...
Derivative works are defined in the Copyright Act to consist of "a work based upon one or more preexisting works, such as a translation, [.], abridgment, condensation, or any other form in which a work may be recast, transformed, or adapted." The definition goes further, however, declaring that "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.'"
[...]
Merely that an LKM may be independently written and conceived fails to remove it from derivative work status: even if it is "an original work of authorship," if it is essentially an "elaboration or other modification," it is still a derivative work of the original Program.
[...]
Torvalds aptly analogizes the relationship of the kernel to a module in terms of the relationship of a novel to a chapter within it. Like a novel, the kernel is made up of many parts, with continuity between them, and a structural relationship holding everything together. In some instances, a chapter setting a scene, or describing an event could stand alone as a short story, and could possibly have even been written before the novel. In other instances, however, a chapter that includes characters, plot points and events referenced elsewhere in the novel could not stand alone. Torvalds sees that kernel modules could be constructed in a sufficiently independent manner that they might not be a legal derivative work of the Linux kernel, but also that many modules —in particular, those with "intimate knowledge about kernel internals"—likely do fall within that boundary.
OK, I think we're starting to agree
"Why are you referring to the GPL as an authority on what constitutes a derivative work? The GPL cannot determine that. Only the copyright statutes and case law can."
My point was that the whether the kernel devs "publicly take the stance" that all code added to linux is derivative doesn't really matter (legally, that is). The GPL is important because it determines how a programmer is allowed to use linux code, and the GPL points us to copyright law as a determinant of whether something is "derivative." So, we agree.
"Merely that an LKM may be independently written and conceived fails to remove it from derivative work status: even if it is "an original work of authorship," if it is essentially an "elaboration or other modification," it is still a derivative work of the original Program."
Sure, but the "elaboration or other modification" part may be very subjective and open to question. By no means is it legally assured that all contributions to linux are necessarily "elaboration or other modifications" (most likely, some are and some aren't). I think you agree, based on what you say later.
Anyway, to get back to the original discussion (regarding whether the kernel can be migrated from GPLv2 to GPLv3 in anything less than an all-or-none fashion), I think that it would not have to be the whole kernel all at once. I think a lot of device drivers could be dual-licensed, with them being used under the GPLv2 when included with the GPLv2 kernel. Many subsystems have been ripped out and totally rewritten over the years, and when that happens the new code could be independent enough to be legitimately dual-licensed. At some point most of the GPLv2-only code will be gone. It just needs to happen on a fairly large-scale basis, as I agree that you could not have lines of dual-licensed code scattered through an otherwise GPLv2 file (for example).
I think I'll stop after this post, but to summarize, I am not convinced that the kernel is legally frozen into GPLv2 forever, just because it doesn't use a "or later" clause. It just boils down to whether Linus and enough other important kernel developers want to go through the effort to make the switch.
Thanks, and no offense has been intended.
Device drivers could
Device drivers could possibly be dual licensed because derived work more weakly applies to them. This isn't about device drivers. This is about people suggesting that Linux could be rewritten incrementally as a magic wand to licensing issues.
With respect to core kernel functionality, it is difficult to understand how a rewritten core kernel component, that is intended to be a drop in replacement for existing functionality and which has a software interface that only works with Linux, could NOT be construed as a derived work of the GPLv2 kernel. Then you're back to the beginning, where you have to get a GPLv3 exception from everyone else.
Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3
"I don't think the GPLv3 is as good a license as v2
...As it is, I don't really see a reason at all [ for GPLv3 ]."
Maybe Linus will see the reason after Microsoft gets through assimilating Linux.
Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3
Maybe Linus will see the reason after Microsoft gets through assimilating Linux.
Maybe, but until then, he kinda thinks the FSF is being a bit like Bush administration, protecting freedom through it's destruction and rhetoric!
This is the FSF we're talking about here, GNU software would be nowhere without GCC. It isn't the comparision (that's pure rhetoric), but that there are no concrete examples of how the GPLv3 might actually take away freedom.
"...Maybe, but until then,
"...Maybe, but until then, he kinda thinks the FSF is being a bit like Bush administration, protecting freedom through it's destruction and rhetoric!"
If that is true, which I doubt, then Linus would be very naive in his world view. FSF has nothing in common with the US Neocons, that comparison is ridiculous. The FSF understands the real problems facing free software in the world today and is not afraid to do what is required, while most people are too naive, uninformed, or misdirected to understand the issues.
I linked to the exact lkml
I linked to the exact lkml email and it's not that hard to find if you read the page, it's in the last paragraph (I never directly attributed words to anyone, I said "he kinda things the FSF is being a bit like"). Regardless, he brought up the Bush administration.
He points out (if you bother to read the whole flame-war) that both, the BSD license advocates and the GPLv3 license advocates claim that their's is the way to more software freedom, and makes the point that the GPLv2 is the correct "tradeoff" so to speak between the two, while being the chosen license for the Linux kernel. I completely agree with Linus Torvalds (and Andrew Morton, and Alan Cox, and others), but I do think we haven't seen all incarnations of DRM yet, and the GPLv3 might not be a bad license to choose in the future if there are no down sides.
It's not my choice though, but just because the FSF sees a different way forward, doesn't mean their license is somehow restricting software freedom. Yes, having copyleft content and open hardware would be the preferred way of getting rid of DRM, but in the meantime, why can't software writers require that their software be buildable from source where ever it runs. I can imagine virtual hardware refusing to run unsigned kernels, and software choosing only to run on "safe" hardware, and the GPLv2 may be enough for it all, but having GPLv3 as a choice can't be bad.
DTrace
Just to clarify - Linux does not have DTrace equivalent. It has something that could became DTrace equivalent in a year or two, but it's not there yet.