I have seen a lot of misinformation posted (and then strongly upvoted) in another thread related to this controversy. I would like to offer a very specific quotation which I believe illustrates the crux of the issue, excerpted from the linked pdf:
Determining copyrightability is thus a fact-specific, case-by-case exercise. In the case of the Linux kernel header files,
most would likely be copyrightable (even with comments removed) because the headers include many macros, inline
functions, and other logic that clearly qualifies as expression. In fact, there are several Linux kernel header files that consist almost entirely of static inline function, which is generally regarded as copyrightable expression. For instance, the Linux kernel header file <sys/byteorder.h> consists of an optimized byteswapping function that was included by the Linux kernel developers because it was clever and well-written – the kind of creativity that constitutes copyright protection. (The initial thread from the Linux kernel mailing list that contributes this particular inline function and
discusses the performance benefit it supplies is available at http://lkml.org/lkml/2002/11/29/68 .) When it created the
“clean” Bionic headers, Google kept this inline function. See http://android.git.kernel.org/?p=platform/bionic.git;a=blob;....
(end quotation)It seems clear these header files do include copyrighted original source code that cannot be stripped of the GPL.
IANAL, but even I can see that it is inaccurate in at least several places:
> Google wanted to use a C library that was not licensed
> under GPLv2 in order to “keep GPL out of user-space.”
> Instead of using glibc, therefore, it created a new C
> library called Bionic.
glibc is licensed under the LGPL; the LGPL does say you have to give users some rights (e.g. to decompile your binaries), but you can link LGPL libraries against code, make a binary, and not provide the binaries. Google could have cut back glibc headers to only include the interface to Linux, licensed the resulting library / header file set under LGPL, and then built a library that uses that LGPLd library, licensed under whatever terms they want (within the minor restrictions of the LGPL), and that would have been entirely within both the spirit and letter of the LGPL.
> But even if you were to undertake this analysis and
> segregate the copyrightable expression from the
> non-copyrightable, it doesn’t change the fact that GPLv2
> covers the entirety of the kernel code and any
> derivatives of it. Picking and choosing pieces of the
> kernel code to include in a new library is creating a
> derivative work, and, accordingly, GPLv2 would still
> apply".
If Google successfully removed all copyrightable code from the Linux headers, and included only the interfaces, then they wouldn't have to comply with the GPL license for the kernel, because the whole point of the doctrine of fair use is that you can do it without a license. It therefore wouldn't need to license its library under the GPL.
If, on the other hand, the doctrine of merger didn't apply, and it wasn't fair use, they would need a license to distribute the kernel headers. The GPL is the only one available, so they would need to comply with its terms. Their arguments that what Google did wasn't fair use seem to be more coherent, although I haven't checked their claimed facts.
The article makes the claim that inline functions and macros in header files are copyrightable. This is plausible (based on what I've heard from lawyers). It then goes on to suggest that applications which use Android's NDK could be declared GPLed if there is a successful challenge to Android's cleaning of header files. This seems far-fetched. Let's work it through. First, Google would have to acknowledge that its headers were GPLed. Presumably this would involve a court case. Brought by who? And on what grounds? Then, somebody, perhaps the winner of the court case, would have to get the application developer to acknowledge that their application was GPLed. This would presumably be another court case. Again, brought by who, and on what grounds? Assuming you did go through these heroic and precedent-setting court cases, you would end up with the source code to a shared library, and you would have to go through it all again to get access to the Java code or other application components which are executed by interpreters. The whole piece reads like Slashdot-esque fear-mongering.
The article then goes on to suggest that Google has "unlocked the kernel from the restrictions of GPL2" and could "create a non-GPL'd fork of the Linux kernel". I can't imagine how they imagine this could possibly be the case, even if some legal challenge determines that they do have non-GPLed headers, and the paper doesn't clarify.
1) This article claims several times that APIs are copyrightable. That is a fairly bold claim to make, and would imply that any libc implementation for linux would need to be licensed by whomever the linux IP holders are. If glibc has not received such a license then it is in violation of the GPL, since it is not released under the GPL.
2) There is the much less strong (and more tenable) claim that functions and macros defined in the headers are copyrightable since they are not a necessary part of interacting with the linux kernel. Google can then become in compliance by removing the offending code and re-implementing it themselves. They could still be liable for damages but...
3) If someone were to sue Google for copyright infringement, they would not be able to compel google to release the code to Bionic under the GPL. That could be part of a settlement, but the most they could hope for in a court is to get Google to stop distributing Bionic, plus receipt of monetary damages.
The point is that implementations are copyrightable and because Google did not strip out macros and inline functions, it has not simply reimplemented an interface but instead created a derived work.
"Google can now use the “clean” Bionic headers to
create a non-GPL’d fork of the Linux kernel,
one that can be extended under proprietary license terms."
No, Google can only write non-GPL'd programs on top of the existing kernel. The kernel itself is GPL'd, regardless of whether the "clean" headers count or not.
I believe the entity that would have to bring any suit against Google would be the Free Software Foundation or their representatives.
However, I don't believe the FSF would have an interest in having an interface be declared copyrightable - just the opposite.
Most GPL suits are extremely forgiving. They give the party the option of removing the code or making it available.
The absolute worst case scenario I could imagine given this is that Google would be forced to make a "cleaner clean-room" implementation of the interface.
The last thing I could imagine the Free Software Foundation wanting is a demonstration that it is impossible to construct proprietary software that runs on Linux.
All this just has the feel of anti-Google FUD...
Edit: yes, gpl-violations.org seems to have done the last Linux I found of but I don't think any Free Software Entity wants to be Oracle's stalking horse against Google.
> The last thing I could imagine the Free Software Foundation wanting is a demonstration that it is impossible to construct proprietary software that runs on Linux.
Not yet, maybe.
Since he started his movement, Richard Stallman made it absolutely clear that he considers proprietary software to be evil in a literal sense of the word. He repeatably said it "should not exists" [1].
Recall that when Richard Stallman introduced the GPL for GNU, there was no LGPL. It was introduced later. Here's the announcement from the GNU bulletin [2]:
"We should by now have finished a new alternative General Public License for certain GNU libraries. This license permits linking the libraries into proprietary executables under certain conditions. The new library license actually represents a strategic retreat. We would prefer to insist as much as possible that programs based on GNU software must themselves be free."
Note also that any GNU system (and therefore Linux) is based on the GCC, distributed under the GPL. It's not just a compiler but also provides a runtime library that most programs need to link to when compiled by GCC. To make this possible, it has a special exception to the GPL. However, since the copyright is completely owned by the FSF, the exception can be revoked, anytime -- making it impossible to compile proprietary software with GCC (unless a fork proves to be viable).
In other words, the FSF has the means and the motive to make proprietary software impossible to run under Linux. All they lack right now, is the political power -- that is, the opportunity to do so.
[random: Gotta love this stuff when its submitted anonymously.]
Long snarky post edited down to the basic soundbite : Copyright is an evolving area where the judiciary is trying hard to keep up.
Google, and other companies, pay people money to create software which, in exchange for that pay (and free munchies etc) those people give complete ownership over to the company. It is completely reasonable to expect Google to do anything reasonable to limit the virality of the GPL's impact on what they do, especially in a product that they charge money for and expect to make serious coin on. Actions I would expect them to pursue are rewrites and selective litigation.
12 comments
[ 4.1 ms ] story [ 41.6 ms ] threadDetermining copyrightability is thus a fact-specific, case-by-case exercise. In the case of the Linux kernel header files, most would likely be copyrightable (even with comments removed) because the headers include many macros, inline functions, and other logic that clearly qualifies as expression. In fact, there are several Linux kernel header files that consist almost entirely of static inline function, which is generally regarded as copyrightable expression. For instance, the Linux kernel header file <sys/byteorder.h> consists of an optimized byteswapping function that was included by the Linux kernel developers because it was clever and well-written – the kind of creativity that constitutes copyright protection. (The initial thread from the Linux kernel mailing list that contributes this particular inline function and discusses the performance benefit it supplies is available at http://lkml.org/lkml/2002/11/29/68 .) When it created the “clean” Bionic headers, Google kept this inline function. See http://android.git.kernel.org/?p=platform/bionic.git;a=blob;....
(end quotation)It seems clear these header files do include copyrighted original source code that cannot be stripped of the GPL.
glibc is licensed under the LGPL; the LGPL does say you have to give users some rights (e.g. to decompile your binaries), but you can link LGPL libraries against code, make a binary, and not provide the binaries. Google could have cut back glibc headers to only include the interface to Linux, licensed the resulting library / header file set under LGPL, and then built a library that uses that LGPLd library, licensed under whatever terms they want (within the minor restrictions of the LGPL), and that would have been entirely within both the spirit and letter of the LGPL.
> But even if you were to undertake this analysis and > segregate the copyrightable expression from the > non-copyrightable, it doesn’t change the fact that GPLv2 > covers the entirety of the kernel code and any > derivatives of it. Picking and choosing pieces of the > kernel code to include in a new library is creating a > derivative work, and, accordingly, GPLv2 would still > apply".
If Google successfully removed all copyrightable code from the Linux headers, and included only the interfaces, then they wouldn't have to comply with the GPL license for the kernel, because the whole point of the doctrine of fair use is that you can do it without a license. It therefore wouldn't need to license its library under the GPL.
If, on the other hand, the doctrine of merger didn't apply, and it wasn't fair use, they would need a license to distribute the kernel headers. The GPL is the only one available, so they would need to comply with its terms. Their arguments that what Google did wasn't fair use seem to be more coherent, although I haven't checked their claimed facts.
The article then goes on to suggest that Google has "unlocked the kernel from the restrictions of GPL2" and could "create a non-GPL'd fork of the Linux kernel". I can't imagine how they imagine this could possibly be the case, even if some legal challenge determines that they do have non-GPLed headers, and the paper doesn't clarify.
God said of remarriage... "more babies".
2) There is the much less strong (and more tenable) claim that functions and macros defined in the headers are copyrightable since they are not a necessary part of interacting with the linux kernel. Google can then become in compliance by removing the offending code and re-implementing it themselves. They could still be liable for damages but...
3) If someone were to sue Google for copyright infringement, they would not be able to compel google to release the code to Bionic under the GPL. That could be part of a settlement, but the most they could hope for in a court is to get Google to stop distributing Bionic, plus receipt of monetary damages.
However, I don't believe the FSF would have an interest in having an interface be declared copyrightable - just the opposite.
Most GPL suits are extremely forgiving. They give the party the option of removing the code or making it available.
The absolute worst case scenario I could imagine given this is that Google would be forced to make a "cleaner clean-room" implementation of the interface.
The last thing I could imagine the Free Software Foundation wanting is a demonstration that it is impossible to construct proprietary software that runs on Linux.
All this just has the feel of anti-Google FUD...
Edit: yes, gpl-violations.org seems to have done the last Linux I found of but I don't think any Free Software Entity wants to be Oracle's stalking horse against Google.
But your main message is right on: Linux kernel developers have no reason to sue Google for what at best is very minor issue.
Not yet, maybe.
Since he started his movement, Richard Stallman made it absolutely clear that he considers proprietary software to be evil in a literal sense of the word. He repeatably said it "should not exists" [1].
Recall that when Richard Stallman introduced the GPL for GNU, there was no LGPL. It was introduced later. Here's the announcement from the GNU bulletin [2]:
"We should by now have finished a new alternative General Public License for certain GNU libraries. This license permits linking the libraries into proprietary executables under certain conditions. The new library license actually represents a strategic retreat. We would prefer to insist as much as possible that programs based on GNU software must themselves be free."
Note also that any GNU system (and therefore Linux) is based on the GCC, distributed under the GPL. It's not just a compiler but also provides a runtime library that most programs need to link to when compiled by GCC. To make this possible, it has a special exception to the GPL. However, since the copyright is completely owned by the FSF, the exception can be revoked, anytime -- making it impossible to compile proprietary software with GCC (unless a fork proves to be viable).
In other words, the FSF has the means and the motive to make proprietary software impossible to run under Linux. All they lack right now, is the political power -- that is, the opportunity to do so.
[1] http://www.networkworld.com/news/2011/031411-richard-stallma... [2] http://www.gnu.org/bulletins/bull10.html#SEC6
Long snarky post edited down to the basic soundbite : Copyright is an evolving area where the judiciary is trying hard to keep up.
Google, and other companies, pay people money to create software which, in exchange for that pay (and free munchies etc) those people give complete ownership over to the company. It is completely reasonable to expect Google to do anything reasonable to limit the virality of the GPL's impact on what they do, especially in a product that they charge money for and expect to make serious coin on. Actions I would expect them to pursue are rewrites and selective litigation.