Skip to contentSkip to conversation

Binary only module overview

DISCUSSION Crispin Cowan 9 participants 30 messages Sep 24, 2001 – Sep 29, 2001

lore
Share & tools
lore
#0 lore raw

Arjan van de Ven <[email protected]> wrote

I'm composing a list of all existing binary only modules, and I got to a list of 26 different modules
... Hewlet Packard - High level security modules (LSM) SGI - High level security modules (LSM) Wirex - High level security modules (LSM)

NOTE: To my knowledge, the above mentioned modules either do not exist or have not been released. The LSM patch itself is entirely GPL'd.

The debate thread http://mail.wirex.com/pipermail/linux-security-module/2001-September/002017.html that Greg KH referred to is about whether LSM (security) modules should ever be permitted to be proprietary. Some feel that all LSM modules should be OSD-compliant Open Source software, while others feel that LSM should continue the existing Linux module policy of permitting proprietary modules only if they do not require changes to the Linux kernel (which would make them a derived work of the kernel).

This post cc'd and reply-to'd to the LSM mailing list.

Crispin

Signature · 4 lines
Crispin Cowan, Ph.D.
Chief Scientist, WireX Communications, Inc. http://wirex.com
Security Hardened Linux Distribution:       http://immunix.org
Available for purchase: http://wirex.com/Products/Immunix/purchase.html
#1 lore raw
Quoted 5 lines

ever be permitted to be proprietary. Some feel that all LSM modules should be OSD-compliant Open Source software, while others feel that LSM should continue the existing Linux module policy of permitting proprietary modules only if they do not require changes to the Linux kernel (which would make them a derived work of the kernel).

With the current lunatic US congress proposals on security, crypto and building big brother into all PC's I'd say allowing non GPL security modules is positively dangerous to the well being of non US citizens

#2 lore raw
Quoted 6 lines · Alan Cox
Quoted 2 lines

ever be permitted to be proprietary. Some feel that all LSM modules should be OSD-compliant Open Source software, while others feel that LSM should continue the existing Linux module policy of permitting proprietary modules only if they do not require changes to the Linux kernel (which would make them a derived work of the kernel).

With the current lunatic US congress proposals on security, crypto and building big brother into all PC's I'd say allowing non GPL security modules is positively dangerous to the well being of non US citizens

Alan made a very interesting point in this post back in June http://lwn.net/2001/0614/a/ac-modules.php3 that the Linux kernel is all fundamentally GPL licensed. Because the kernel is a composite work of many authors, and all that code was contributed under the GPL 2. It would require the unanimous consent of all the copyright holders to change the license.

That it is GPL licensed in turn has implications. One of them is that you are not allowed to impose additional constraints on distribution:

* Clause 4: "You may not copy, modify, sublicense, or distribute the Program except as expressly provided under this License."
* Clause 6: "... You may not impose any further restrictions on the recipients' exercise of the rights granted herein."

Therefore, any additional constraints people may wish to impose, such as Greg's comment in security.h, are invalid. When someone receives a copy of the Linux kernel, the license is pure, vanilla GPL, with no funny riders.*

The question of whether proprietary (non-GPL) modules are permitted is a matter of opinion. As Alan states in the June post above, Linus has given his opinion (that binary modules are ok, so long as it doesn't require kernel changes to run) but that is *only* Linus' opinion. Others may have different opinions, but they are all just opinions until the courts eventually rule on how the GPL is to be interpreted in this matter.

In light of all that, I propose that the LSM project take a strictly neutral stance on the question of binary modules. LSM imposes no new restrictions (which would be invalid anyway) and makes no judgment on whether binary modules are appropriate. As such, we would replace Greg's comment in security.h (and in all other LSM-specific files) with a comment that says "Copyright 2001 <authors>, Licensed under the GPL. See the Linux Kernels COPYING file for details."

How does that sound to folks?

Crispin

[*] The singular exception is the rider that Linus prepended to the Linux COPYING file, scoping what the GPL applies to. Presumably this rider was added before multiple authors got involved. If you wanna challenge Linus's exception and insist that all Linux applications are GPL'd, that's another thread :-)

Signature · 4 lines
Crispin Cowan, Ph.D.
Chief Scientist, WireX Communications, Inc. http://wirex.com
Security Hardened Linux Distribution:       http://immunix.org
Available for purchase: http://wirex.com/Products/Immunix/purchase.html
#3 lore raw
Quoted 1 line · Crispin Cowan

How does that sound to folks?

It sounds both wise and consistant with what I expected from LSM. I can live with that.

Quoted 3 lines

Crispin

...

Quoted 5 lines
Signature · 4 lines
Crispin Cowan, Ph.D.
Chief Scientist, WireX Communications, Inc. http://wirex.com
Security Hardened Linux Distribution:       http://immunix.org
Available for purchase: http://wirex.com/Products/Immunix/purchase.html

Bravo,
J. Melvin Jones

|>------------------------------------------------------
||  J. MELVIN JONES            [email protected] 
|>------------------------------------------------------
||  Microcomputer Systems Consultant  
||  Software Developer
||  Web Site Design, Hosting, and Administration
||  Network and Systems Administration
|>------------------------------------------------------
||  http://www.jmjones.com/
|>------------------------------------------------------
#5 lore raw
In reply to Alan Cox
Quoted 5 lines · Crispin Cowan

Therefore, any additional constraints people may wish to impose, such as Greg's comment in security.h, are invalid. When someone receives a copy of the Linux kernel, the license is pure, vanilla GPL, with no funny riders.*

My comment in security.h that I proposed [1] does not add any additional constraints to the license that is currently on the file. All it does is explicitly state the licensing terms of it, so that there shall be no confusion regarding it's inclusion in programs. If you think this is adding an additional restriction to the file, please explain.

If you were to include a GPL licensed user space header file in a closed source program, of course you would be violating that license. So why do people think that since a file is in include/linux that the license attached to that file is no longer valid?

Yes it is true that a variety of companies currently ship binary modules for Linux. And hopefully in the compilation of those modules they do not include any GPL licensed header files. I know some companies go to great lengths to prevent this from happening.

thanks,

greg k-h

[1] Included here for those who did not see it on the linux-security-module mailing list:

This file may not be included in any code not licensed under the list of accepted free software licenses as defined in module.h contained in this same directory.

#4 lore raw
In reply to Crispin Cowan
Quoted 7 lines

* Clause 6: "... You may not impose any further restrictions on the recipients' exercise of the rights granted herein."

Therefore, any additional constraints people may wish to impose, such as Greg's comment in security.h, are invalid. When someone receives a copy of the Linux kernel, the license is pure, vanilla GPL, with no funny riders.*

Greg's comment states what is already stated by the GPL. It might not suit your personal desire, but thats tough. The matter is one of "linking" - in a legal not a technical sense and of what is and is not a derivative work. If your code is a derivative work of GPL code then the GPL is totally clear on the matter.

Alan

#6 lore raw
In reply to Alan Cox
Quoted 11 lines · Greg KH
Quoted 2 lines · Crispin Cowan

Therefore, any additional constraints people may wish to impose, such as Greg's comment in security.h, are invalid. When someone receives a copy of the Linux kernel, the license is pure, vanilla GPL, with no funny riders.*

My comment in security.h that I proposed [1] does not add any additional constraints to the license that is currently on the file. All it does is explicitly state the licensing terms of it, so that there shall be no confusion regarding it's inclusion in programs. If you think this is adding an additional restriction to the file, please explain.

What your comment does is explicitly state your *interpretation* of the implications of the GPL. As is manifestly obvious, the GPL is subject to lots of interpretation, especially in the area of what is a "derived work." We are on the safest legal ground if we simply state that the file in question is GPL'd, and leave it at that.

Quoted 3 lines

If you were to include a GPL licensed user space header file in a closed source program, of course you would be violating that license.

That is not clear to me. I have been unable to find a definitive reference that states that is the case. If so, it is problematic, because then every user-land program that ever #include'd errno.h from glibc is GPL'd, because glibc #include's errno.h, among other GPL'd kernel header files. Are you sure you want to declare nearly all proprietary Linux applications to be in violation of the GPL?

We have a Schrodinger's Cat problem of whether the courts will eventually rule that modules are derivative works of the kernel. There are two cases here. Either:

* Binary modules are permitted by the kernel's GPL: if this is the case, then Greg's comment is invalid, and misleading.
* Binary modules are not permitted by the kernel's GPL: if this is the case, then Greg's comment is redundant, and just marking the file "GPL" is sufficient.

IMHO, in neither case is the special language appropriate. This file is GPL'd, and we should stop playing lawyer by trying to interpret what that means.

If you (Greg, Alan) are confident that your interpretation of the GPL is correct, then just marking the files as GPL should be sufficient. What purpose is served by saying anything else?

Crispin

Signature · 4 lines
Crispin Cowan, Ph.D.
Chief Scientist, WireX Communications, Inc. http://wirex.com
Security Hardened Linux Distribution:       http://immunix.org
Available for purchase: http://wirex.com/Products/Immunix/purchase.html
#7 lore raw
Quoted 3 lines

If you (Greg, Alan) are confident that your interpretation of the GPL is correct, then just marking the files as GPL should be sufficient. What purpose is served by saying anything else?

It avoids confusion with some of Linus' more strange interpretations which as a major copyright holder may otherwise be considered to the detriment of other copyright holders

#8 lore raw
In reply to Alan Cox
Quoted 11 lines · Crispin Cowan
Quoted 3 lines · Greg KH

If you were to include a GPL licensed user space header file in a closed source program, of course you would be violating that license.

That is not clear to me. I have been unable to find a definitive reference that states that is the case. If so, it is problematic, because then every user-land program that ever #include'd errno.h from glibc is GPL'd, because glibc #include's errno.h, among other GPL'd kernel header files. Are you sure you want to declare nearly all proprietary Linux applications to be in violation of the GPL?

That is an issue to take up with the glibc authors, not me. If something like this bothers you, then use a libc that does not include kernel header files (which has been pointed out by the kernel authors that they should not be doing.) dietLibc [1] works for me quite well and does not contain any kernel header files.

Quoted 3 lines

If you (Greg, Alan) are confident that your interpretation of the GPL is correct, then just marking the files as GPL should be sufficient. What purpose is served by saying anything else?

As Alan stated, to reduce confusion as to the wishes of the copyright holders of the file. 3 out of the 4 current holders agree to this wording, and as the dissenting party, you are free to disagree and keep that wording from the file (however a small note in it detailing this disagreement might be a nice thing to do.)

thanks,

greg k-h

[1] http://www.fefe.de/dietlibc/

#9 lore raw
In reply to Alan Cox
Quoted 9 lines · Crispin

We have a Schrodinger's Cat problem of whether the courts will eventually rule that modules are derivative works of the kernel. There are two cases here. Either:

* Binary modules are permitted by the kernel's GPL: if this is the case, then Greg's comment is invalid, and misleading.
* Binary modules are not permitted by the kernel's GPL: if this is the case, then Greg's comment is redundant, and just marking the file "GPL" is sufficient.

This is the most concise explanation of the problem so far.

I don't like binary security modules, and I won't use them nor write them, because that's my desicion to make. Whether they are legal or not is not. It's all there in the GPL for well paid lawyers to determine. It doesn't seem that there's any benefit to adding to the existing licensing terms -- it all boils down to which of the two above cases is true.

Quoted 7 lines

IMHO, in neither case is the special language appropriate. This file is GPL'd, and we should stop playing lawyer by trying to interpret what that means.

If you (Greg, Alan) are confident that your interpretation of the GPL is correct, then just marking the files as GPL should be sufficient. What purpose is served by saying anything else?

Ding.

I wouldn't mind any 'We heartily suggest/prefer' wording, just let's not get into anything legal or restrictive or the license could become internally contradictory and make it more difficult to prosecute GPL infractions. (Where there are two different ideas in this group about what constitutes an infraction...)

I always assumed that the LSM was to be treated no differently than existing modules -- LKMs may be close source. It was never said otherwise anywhere when this project was formed. Changing (ahem, clarifying) it at this stage of the game is a bad thing.

Had I been able to contribute actual code to this project (the days never are long enough, are they) I'd have more of a direct say. Taking an approach like 'oh, X of Y developers say one thing, now that we asked' is not a fair way to make desicions. Though many on this list have not contributed code, they have made comments and suggestions which were used by the primary coders. Perhaps we should take a grand survey of everyone and see what they think, since they all had some input, if not output, into the code as it stands? No, that's even worse.

The LSM code is GPLd. That was a stated requirement from day one. Nothing else was. Leave the interpretation of binary module acceptibility/nonacceptibility to the GPL and the lawyers, that's not our job.

--
Brian Hatch e ^ (i*pi) = -1
Systems and
Security Engineer
www.hackinglinuxexposed.com

Every message PGP signed

#10 lore raw
In reply to Alan Cox

On Wed, Sep 26, 2001 at 12:17:37PM -0700, Crispin Cowan wrote:

[snip]

Quoted 6 lines

That is not clear to me. I have been unable to find a definitive reference that states that is the case. If so, it is problematic, because then every user-land program that ever #include'd errno.h from glibc is GPL'd, because glibc #include's errno.h, among other GPL'd kernel header files. Are you sure you want to declare nearly all proprietary Linux applications to be in violation of the GPL?

AFAIK, the glibc (and most other libraries) are LGPL rather than GPL.

[snip]

/David Weinehall
_ _
// David Weinehall <[email protected]> /> Northern lights wander \\

//  Project MCA Linux hacker        //  Dance across the winter sky //
\>  http://www.acc.umu.se/~tao/    </   Full colour fire           </
#11 lore raw
Quoted 10 lines · David Weinehall
Quoted 6 lines · Crispin Cowan

That is not clear to me. I have been unable to find a definitive reference that states that is the case. If so, it is problematic, because then every user-land program that ever #include'd errno.h from glibc is GPL'd, because glibc #include's errno.h, among other GPL'd kernel header files. Are you sure you want to declare nearly all proprietary Linux applications to be in violation of the GPL?

AFAIK, the glibc (and most other libraries) are LGPL rather than GPL.

What about programs that include header files from /usr/include/linux, /usr/include/asm, and/or /usr/include/scsi?

Signature · 1 line
Ignacio Vazquez-Abrams  <[email protected]>
#12 lore raw
Quoted 2 lines

What about programs that include header files from /usr/include/linux, /usr/include/asm, and/or /usr/include/scsi?

I believe you cannot copyright an interface just an implementation of it. I suspect someone more familiar in law can give the required precise info on that boundary

That is the security layer issue is one of "does it depend on the linux kernel to work, is it deriving from the kernel and the GPL'd module for security plugins" not about the precise structs and #defines.

Given the SSSCA we have to be very clear on this issue, and if its not clear I might be best to kill the entire uncertainty by not including the LSM patch in Linux until the US government returns to sanity

#13 lore raw
Quoted 10 lines · Alan Cox
Quoted 2 lines

What about programs that include header files from /usr/include/linux, /usr/include/asm, and/or /usr/include/scsi?

I believe you cannot copyright an interface just an implementation of it. I suspect someone more familiar in law can give the required precise info on that boundary

That is the security layer issue is one of "does it depend on the linux kernel to work, is it deriving from the kernel and the GPL'd module for security plugins" not about the precise structs and #defines.

There are two things that work together cause problems, however:

1) The kernel header files are part of the Linux package, therefore unless a specific license is applied to these files, they are covered by the license applied to the entire package, i.e., GPL.

2) I agree that the structs and #define values contained within the header files are not a problem; there is no transfer of code, only the information about data formats and contents. OTOH, the #define macros, e.g., the ioctl() stuff, may cause a problem due to the fact that cpp does in fact take the code contained in these macros and inserts it into the program.

Therefore, it can be said that anything that uses the header macros must also be under the GPL. Certainly this is a ridiculous idea, but it is one that requires additional clarification.

The best solution is probably to make all header files in <linux-src>/include LGPLed, and make the additional requirement in the Linux license that anything in <linux-src>/include must also be LGPLed.

Signature · 1 line
Ignacio Vazquez-Abrams  <[email protected]>
#24 lore raw
In reply to Alan Cox
Quoted 6 lines · David Weinehall
Quoted 2 lines · Crispin Cowan

That is not clear to me. I have been unable to find a definitive reference that states that is the case. If so, it is problematic, because then every user-land program that ever #include'd errno.h from glibc is GPL'd, because glibc #include's errno.h, among other GPL'd kernel header files. Are you sure you want to declare nearly all proprietary Linux applications to be in violation of the GPL?

AFAIK, the glibc (and most other libraries) are LGPL rather than GPL.

It appears that while glibc is LGPL, it in turn #include's stuff from the kernel. It more or less has to; otherwise glibc has to guess the format of data structures the kernel is going to export.

Greg is partially correct that this is a licensing issue that the glibc maintainers need to resolve. However, I am not convinced that they can resolve it on their own. I see only the following possible resolutions:

* we all decide (an opinion) that #include some_gpl.h does not GPL the code doing the including
* glibc changes its license to GPL, which would make it unpopular among proprietary application developers
* Linux maintainers decide to change the license on the relevant header files to LGPL

If one of the above does not happen, then I think I can derive "false" :-)

Crispin

Signature · 4 lines
Crispin Cowan, Ph.D.
Chief Scientist, WireX Communications, Inc. http://wirex.com
Security Hardened Linux Distribution:       http://immunix.org
Available for purchase: http://wirex.com/Products/Immunix/purchase.html
#14 lore raw
In reply to Ignacio Vazquez-Abrams
Quoted 3 lines

The best solution is probably to make all header files in <linux-src>/include LGPLed, and make the additional requirement in the Linux license that anything in <linux-src>/include must also be LGPLed.

If you want to chase down each author of each file and lobby them indvidually go ahead.

Alan

#25 lore raw
In reply to Alan Cox
Quoted 11 lines · Greg KH
Quoted 7 lines · Crispin Cowan
Quoted 3 lines · Greg KH

If you were to include a GPL licensed user space header file in a closed source program, of course you would be violating that license.

That is not clear to me. [... #include glibc -> #include kernel]

That is an issue to take up with the glibc authors, not me.

Fair enough. I pointed it out to make clear that there are licensing problems and ambiguities if one strictly insists that #include some_gpl.h implies that the code is GPL'd. If that is the case, then there are MUCH bigger licensing problems than LSM, so don't take it up with me, either :-)

Quoted 5 lines
Quoted 2 lines

If you (Greg, Alan) are confident that your interpretation of the GPL is correct, then just marking the files as GPL should be sufficient. What purpose is served by saying anything else?

As Alan stated, to reduce confusion as to the wishes of the copyright holders of the file.

However, it is an outright contradiction to some wishes of some other copyright holders of the kernel (Linus' binary kernel opinion). Since revisions to the kernel's GPL are explicitly prohibited, it seems to me that this statement of one side of the controversy as if it were a fact increases confusion rather than decreases it.

Quoted 2 lines

a small note in it detailing this disagreement might be a nice thing to do.

Fair enough. How about this:

"This file is GPL. See the Linux Kernel's COPYING file for details. There is controversy over whether this permits you to write a module that #includes this file without placing your module under the GPL. Consult your lawyer for advice."

I'm really trying to be constructive here. There is a real licensing problem over whether binary modules are legitimate at all, and the issue is not special to LSM. I'm trying to get LSM out of the way so that the advocates of either side can fight it out without smushing LSM in the middle :-)

Crispin

Signature · 4 lines
Crispin Cowan, Ph.D.
Chief Scientist, WireX Communications, Inc. http://wirex.com
Security Hardened Linux Distribution:       http://immunix.org
Available for purchase: http://wirex.com/Products/Immunix/purchase.html
#15 lore raw
In reply to Alan Cox
Quoted 6 lines · Alan Cox
Quoted 3 lines

The best solution is probably to make all header files in <linux-src>/include LGPLed, and make the additional requirement in the Linux license that anything in <linux-src>/include must also be LGPLed.

If you want to chase down each author of each file and lobby them indvidually go ahead.

Is there an official place where the responses from them should be collected or should I just handle them and notify the list when I receive them and what they are?

Signature · 1 line
Ignacio Vazquez-Abrams  <[email protected]>
#26 lore raw
In reply to Crispin Cowan
Quoted 5 lines

I'm really trying to be constructive here. There is a real licensing problem over whether binary modules are legitimate at all, and the issue is not special to LSM. I'm trying to get LSM out of the way so that the advocates of either side can fight it out without smushing LSM in the middle :-)

Yes - I agree. The question is "can you be using the LSM module" not the headers - since LSM is GPL and your work relies on it

Alan

#29 lore raw
In reply to Alan Cox
Quoted 8 lines · Crispin Cowan
Quoted 2 lines

a small note in it detailing this disagreement might be a nice thing to do.

Fair enough. How about this:

"This file is GPL. See the Linux Kernel's COPYING file for details. There is controversy over whether this permits you to write a module that #includes this file without placing your module under the GPL. Consult your lawyer for advice."

That's acceptable to me. I'll go add that to the tree.

thanks,

greg k-h

#16 lore raw
In reply to Ignacio Vazquez-Abrams
Quoted 4 lines · Alan Cox

Given the SSSCA we have to be very clear on this issue, and if its not clear I might be best to kill the entire uncertainty by not including the LSM patch in Linux until the US government returns to sanity

Or place some kind of markings on the symbols/functions that the LSM code exports stating that these symbols/functions can only be called from GPL licensed code.

But to do that, it would require the cooperation of all of the LSM authors :)

thanks,

greg k-h

#27 lore raw
In reply to Alan Cox
Quoted 5 lines · Alan Cox
Quoted 2 lines

I'm really trying to be constructive here. There is a real licensing problem over whether binary modules are legitimate at all, and the issue is not special to LSM. I'm trying to get LSM out of the way so that the advocates of either side can fight it out without smushing LSM in the middle :-)

Yes - I agree. The question is "can you be using the LSM module" not the headers - since LSM is GPL and your work relies on it

I'm confused as to which question Alan is asking, so I'll answer several :-)

* "What is the licensing on the LSM module?" Crispin's pedantic
response: there is no "the" LSM module. LSM is a patch to enrich the existing Linux loadable kernel module interface to allow access control modules. These modules do exist:
* There is a dummy module (useful as a template) which just does the superuser check ("UID==0" and access is granted) and it is GPL'd.
* There is a module that embodies the guts of POSIX capabilities, and it is GPL'd.
* There is a port of SELinux to LSM (produced by the SELinux team) and it is GPL'd.
* "Can you use the LSM interface with a non-GPL module, if you eschew GPL'd .h files?" Crispin's opinion: I dunnow, but the answer is the same as the answer for current Linux loadble kernel modules. If loading a module is "linking", then all modules & device drivers must be GPL'd.
* "Can a non-GPL LSM module #include a GPL'd .h?" Crispin's opinion:
I dunnow, but the answer is the same as the answer for many current Linux applications. I know, this is the question Alan did not ask :-)

Crispin is not a lawyer/judge, so his opinion doesn't matter anyway :-)

Crispin

Signature · 4 lines
Crispin Cowan, Ph.D.
Chief Scientist, WireX Communications, Inc. http://wirex.com
Security Hardened Linux Distribution:       http://immunix.org
Available for purchase: http://wirex.com/Products/Immunix/purchase.html
#28 lore raw
In reply to Alan Cox
Quoted 8 lines · Alan Cox
Quoted 5 lines

I'm really trying to be constructive here. There is a real licensing problem over whether binary modules are legitimate at all, and the issue is not special to LSM. I'm trying to get LSM out of the way so that the advocates of either side can fight it out without smushing LSM in the middle :-)

Yes - I agree. The question is "can you be using the LSM module" not the headers - since LSM is GPL and your work relies on it

Unfortunately, the commentary in /usr/src/linux/COPYING exempting programs that use the syscall interface is clear as mud.

I can read it as saying "the syscall interface is hereby granted an exemption, and other normal uses of the kernel are specifically NOT exempted". In that case, all authors of closed-source loadable modules are heretics and need to be burnt at the stake. ;)

I can equally easily read it as "normal use of the kernel does not fall under 'derivative work', and the syscall inferface is cited as one example of normal use". Given that Linus has been quoted elsewhere as saying that closed-source modules are not *inherently* evil, this may be the intended reading.

I can equally easily read it as "Linus wrote it when the syscall interface WAS the only interface, and never updated it for loadable modules...." ;)

Personally, I'd not be at all surprised to find out that none of my 3 readings are anywhere close to the actual meaning as decided by a court of law....

Valdis Kletnieks Operating Systems Analyst Virginia Tech

#17 lore raw
In reply to Greg KH
Quoted 3 lines

Or place some kind of markings on the symbols/functions that the LSM code exports stating that these symbols/functions can only be called from GPL licensed code.

GPL_EXPORT_SYM is coming, has been discussed and tentatively agreed upon so that we can for example have libraries of GPL code that are GPL module only usable, while still exporting clear interfaces for nonfree users when appropriate (eg device drivers)

Turning existing in kernel exports GPL_ all of a sudden is not going to happen.

#18 lore raw
In reply to Greg KH
Quoted 4 lines · Alan Cox

GPL_EXPORT_SYM is coming, has been discussed and tentatively agreed upon so that we can for example have libraries of GPL code that are GPL module only usable, while still exporting clear interfaces for nonfree users when appropriate (eg device drivers)

So in your opinion, would making the LSM public functions use GPL_EXPORT_SYM (when it is available) increase the odds of the patch being accepted into the kernel?

thanks,

greg k-h

#19 lore raw
Quoted 3 lines

So in your opinion, would making the LSM public functions use GPL_EXPORT_SYM (when it is available) increase the odds of the patch being accepted into the kernel?

Not sure - thats a Linus question

#20 lore raw

It has been implied that I am not being "politically correct" on this list. Yeah, I know. That's only one realm of the many ways to be correct. Read on...

--------

I have been reading everything I can access about GPL in the last few days, and, while I'm no snake[1], it seems to me that everything Greg's additional "restriction" adds is probably already covered in the GPL. It also seems to my somewhat-logical-but-not-legal-mind that Linus MAY be stretching the GPL if he explicitly allows binary modules.

However, the way I understand copyright laws (admittedly "vague"): it is up to the copyright holder to defend his product, not anybody else. If Linus, or anybody else, wants to say "okay, I won't file suit against people who abuse me in this manner ... XYZ", that's his/her right. It's less restrictive, and (it would seem) compatable to GPL, IMHO. No government agency is going to SWOOP in and challenge his right to "fail to prosecute."

The way this percolates down, in my mind, is that GPL is a pretty darn good document. If LSM just GPLs it's code, it can still decide (imho), in the ultimate event, to pursue any violations of its licensing. How the courts may act on the "legacy" of the licensor above... is up to the court.

I think it's perfectly acceptable to implement any *legal-technical* means to hide your interface that you choose (data isn't covered by GPL, can somebody create a data-based key that stops GPL'd code from working? Um... maaaaybeeee (^_^). Is a module "code" or "data"? maaaaaaaaaaaybe) I also think, from my reading, that you are violating the GPL if you add other "special restrictions."

Suppose I add "You cannot copy or distribute this portion of the source to anybody for any reason." to the license? Would that be legal? Strictly NOT is my reading. It's restrictive and, therefore, not "compatible." The proposed few lines MAY someday be interpretted as such. *BANG*, you're on legal-quicksand.

Greg (et al), I have great respect for your opinion... but you just may be breaking GPL by adding a "further restriction." If you're not adding a "further restriction", leave it out... it's already there. If some court somewhere finds that this is not part of the GPL restriction, you break from GPL and ... well, then it's anybody's game.

Get out your pencil, draw the logic diagram... you're making a logical mistake.

Um, and logic and law are not necessarily compatible. :)

Trust GPL,
J. Melvin Jones

P.S. -- I've got an interface for similar purposes spec'd and partly written. It is NOT efficient at doing what LSM does, but IS more flexible and allows more things to be done (stipulation). For certain purposes (mine), it may actually be faster, but IN GENERAL, the special case I'm addressing is NOT similar to LSM. It might be better for my and some other solutions. Majority solutions? Who knows? Admittedly, I now see problems with implementation that were cited before when some of my ideas were "trampled." I'm tackling them, not dismissing them.

No worries, I don't have even the REMOTEST desire to be included in the Kernel... I just want something that works and enhances Linux Security.

That being said, I have been advised and believe that I will HAVE to GPL the kernel-side patches, the code, and everything "south" of the API, and gladly will so do. I will also release (least-restrictively-as-possible) any parts necessary to connect to my MPI (module programming interface.) Anything on the OUTSIDE of my interface belongs to whomever created it IMHO. Let the court decide how Linus' and anybody else's interest should be protected, IF they sue... (and, I think I'm doing the RIGHT thing, so I sleep at night and don't worry) ... so be it. If they can build a better mousetrap... hey, KEWL!!!!

Is this a political thing instead of a technical thing? I think it may be, at least for Greg (et al.) Make Linux better and you can sleep at night. Don't let somebody make it worse: you are honorable. But where does LSM fit into that? I'd like to see it make things better, AND open to non-open solutions for making it better. GPL gives you, imh-nonlegal-opinion, the full right to band the Copyright holders together to pursue "legal recoarse" to fight abuse. REAL ABUSE, not just some political ideal.

If you can't sleep at night because somebody else can use your code for purposes OTHER than you envisioned... you shouldn't be writing Open Software. Isn't that the real purpose... to let the next guy stand on your shoulders?

--------

[1] SNAKE -- the best sort of thing to hire when your garden is full of rats. A "SNAKE" is a REPTILE, a term which is also euphemistically
used, by J. Melvin Jones, to describe Insurance Agents, CPA's, and any other people who "bridge the gap between the logical, the ethical, and the legal." No disrespect intended, EVER. Sometimes COLD BLOODEDNESS is a very useful attribute for consultants to mammals.

|>------------------------------------------------------
||  J. MELVIN JONES            [email protected] 
|>------------------------------------------------------
||  Microcomputer Systems Consultant  
||  Software Developer
||  Web Site Design, Hosting, and Administration
||  Network and Systems Administration
|>------------------------------------------------------
||  http://www.jmjones.com/
|>------------------------------------------------------
#21 lore raw
Quoted 5 lines · On Fri, 28 Sep 2001 [email protected] wrote:

However, the way I understand copyright laws (admittedly "vague"): it is up to the copyright holder to defend his product, not anybody else. If Linus, or anybody else, wants to say "okay, I won't file suit against people who abuse me in this manner ... XYZ", that's his/her right.

But could he decide that for the code of one of the other few-hundred copyright holders in the kernel ?

Note that even if Linus says he thinks something is ok, people would still need to respect the rights of all the other copyright holders to the kernel.

Interesting thing is that Richard Stallman never seemed willing to take this leap and trust the GPL, but instead requires people to sign over their copyright to the FSF before code is accepted into a GNU project ;))

cheers,

Rik

Signature · 5 lines
IA64: a worthy successor to i860.

http://www.surriel.com/		http://distro.conectiva.com/

Send all your spam to [email protected] (spam digging piggy)
#22 lore raw
Quoted 11 lines · Rik van Riel
Quoted 5 lines · On Fri, 28 Sep 2001 [email protected] wrote:

However, the way I understand copyright laws (admittedly "vague"): it is up to the copyright holder to defend his product, not anybody else. If Linus, or anybody else, wants to say "okay, I won't file suit against people who abuse me in this manner ... XYZ", that's his/her right.

But could he decide that for the code of one of the other few-hundred copyright holders in the kernel ?

Quoted 8 lines

Note that even if Linus says he thinks something is ok, people would still need to respect the rights of all the other copyright holders to the kernel.

Interesting thing is that Richard Stallman never seemed willing to take this leap and trust the GPL, but instead requires people to sign over their copyright to the FSF before code is accepted into a GNU project ;))

Geeeez, I'm no snake, but my belief would be NO, except that anyone who is operating under the license of another can't usually exceed that license. On the other hand, he couldn't grant "special license" anyway.

It's a pretty interesting legal issue, but GPL seems, to me, to cover all the avenues AND pretty effectively prohibits "more restrictive" licensing. I have NO idea if that's "legal", but, if it falls in court, there may be many other parts of GPL that fall at the same time. And, hey, if MORE restrictive licenses are allowed... GPL is full of holes the size of trucks. GPL is excellent... I wouldn't TOUCH it unless I expected to be laughed out of court.

In my experience, contracts restrict. Anything not prohibitted is allowed, except where EVERYTHING is prohibitted and a specific list of things are allowed. GPL seems to use that: to allow nothing but what is mentioned, and even imposes conditions on that. What little training in contract law (Insurance and Securities Industries) I have says either are possible. The wording of GPL is about as tight as any contract as I have ever read.

In essence: GPL prohibits ALL rights except where granted. Ergo, prohibitting another right is superfluous and dangerous within the construction of the document.

Adding ANYTHING, even a comma, seems dangerous. Linus can release his interest, but I don't know if he can release ANYBODY ELSE's interest.

OPINION: STICK WITH GPL,
J. Melvin Jones

Quoted 11 lines

cheers,

Rik

Signature · 5 lines
IA64: a worthy successor to i860.

http://www.surriel.com/		http://distro.conectiva.com/

Send all your spam to [email protected] (spam digging piggy)
#23 lore raw
In reply to [email protected]
Quoted 4 lines

If you can't sleep at night because somebody else can use your code for purposes OTHER than you envisioned... you shouldn't be writing Open Software. Isn't that the real purpose... to let the next guy stand on your shoulders?

Thats the BSD license. The GPL is where you get to stand on each others shoulders and goon style climb over any future walls together

Alan