Skip to contentSkip to conversation

Module Licensing?

DISCUSSION Kevin D. Wooten 20 participants 45 messages Oct 30, 2001 – Nov 9, 2001

lore
Share & tools
lore
#0 lore raw

After reading the posts about the MODULE_LICENSE macro, I am in disbelief. I was under the impression that one could write a "closed-source" module and distribute it in binary form, and be in compliance. Please tell me I am wrong? We use Linux as a platform for some data acquisition, and we are currently distributing ( in very limited quantity to people who would already have signed an NDA ) modules that currently have no official license as yet. We are researching which license to use, but according to these post's we have almost no choice, Open Source or not at all!

-kw

#1 lore raw
Quoted 9 lines · Kevin D. Wooten

After reading the posts about the MODULE_LICENSE macro, I am in disbelief. I was under the impression that one could write a "closed-source" module and distribute it in binary form, and be in compliance. Please tell me I am wrong? We use Linux as a platform for some data acquisition, and we are currently distributing ( in very limited quantity to people who would already have signed an NDA ) modules that currently have no official license as yet. We are researching which license to use, but according to these post's we have almost no choice, Open Source or not at all!

No, you just can't use certain symbols if you're not GPL. If your
code already works, then you're fine, as previously existing symbols
will not be thus restricted... You can just make your MODULE_LICENSE == "mine-all-mine...including-all-my-bugs"

Ben

Quoted 7 lines
(quoted diff) +0 -2
-kw
-

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 3 lines
Ben Greear <[email protected]>       <Ben_Greear AT excite.com>
President of Candela Technologies Inc      http://www.candelatech.com
ScryMUD:  http://scry.wanfear.com     http://scry.wanfear.com/~greear
#2 lore raw
In reply to Kevin D. Wooten
Quoted 12 lines · Ben Greear
Quoted 9 lines · Kevin D. Wooten

After reading the posts about the MODULE_LICENSE macro, I am in disbelief. I was under the impression that one could write a "closed-source" module and distribute it in binary form, and be in compliance. Please tell me I am wrong? We use Linux as a platform for some data acquisition, and we are currently distributing ( in very limited quantity to people who would already have signed an NDA ) modules that currently have no official license as yet. We are researching which license to use, but according to these post's we have almost no choice, Open Source or not at all!

Quoted 3 lines

No, you just can't use certain symbols if you're not GPL. If your
code already works, then you're fine, as previously existing symbols
will not be thus restricted... You can just make your MODULE_LICENSE == "mine-all-mine...including-all-my-bugs"

Ugghh! Don't confuse/equate MODULE_LICENSE with EXPORT_SYMBOL_GPL_ONLY; two different animals, two differnet goals. See archives for more info.

Quoted 6 lines

Ben

Quoted 3 lines
(quoted diff) +0 -2
-kw
-
Signature · 4 lines
===============
-- TimO
--------------------==============++==============--------------------
                             No Cool .sig
#3 lore raw
In reply to Kevin D. Wooten
Quoted 14 lines · TimO
Quoted 12 lines · Ben Greear
Quoted 9 lines · Kevin D. Wooten

After reading the posts about the MODULE_LICENSE macro, I am in disbelief. I was under the impression that one could write a "closed-source" module and distribute it in binary form, and be in compliance. Please tell me I am wrong? We use Linux as a platform for some data acquisition, and we are currently distributing ( in very limited quantity to people who would already have signed an NDA ) modules that currently have no official license as yet. We are researching which license to use, but according to these post's we have almost no choice, Open Source or not at all!

Quoted 6 lines
Quoted 3 lines

No, you just can't use certain symbols if you're not GPL. If your
code already works, then you're fine, as previously existing symbols
will not be thus restricted... You can just make your MODULE_LICENSE == "mine-all-mine...including-all-my-bugs"

Ugghh! Don't confuse/equate MODULE_LICENSE with EXPORT_SYMBOL_GPL_ONLY; two different animals, two differnet goals. See archives for more info.

Ok, but regardless of the politics, if you're not GPL as was the original poster's issued, then you will not be able to use the GPL_ONLY symbols, right? They aren't the same, but they are inter-related, unless I completely mis-understand....

Ben

Signature · 3 lines
Ben Greear <[email protected]>       <Ben_Greear AT excite.com>
President of Candela Technologies Inc      http://www.candelatech.com
ScryMUD:  http://scry.wanfear.com     http://scry.wanfear.com/~greear
#4 lore raw
In reply to Kevin D. Wooten
Quoted 3 lines

After reading the posts about the MODULE_LICENSE macro, I am in disbelief. I was under the impression that one could write a "closed-source" module and distribute it in binary form, and be in compliance. Please tell me I am

Providing your work constitutes a seperate work that isn't a derivative work. Thats a lawyer definition and one for lawyers

Quoted 5 lines

wrong? We use Linux as a platform for some data acquisition, and we are currently distributing ( in very limited quantity to people who would already have signed an NDA ) modules that currently have no official license as yet. We are researching which license to use, but according to these post's we have almost no choice, Open Source or not at all!

The tag is purely so we can tell free stuff apart. If you tag a module

MODULE_LICENSE("(c) Copyright EvilCorp, All rights wronged");

it will still load but the user will be advised not to report bugs to the kernel lists and the like. The oops data will similarly be marked "tainted" so that we can discard it.

It's a filter so we don't have to handle bugs in your driver, nothing more.

#5 lore raw
In reply to Kevin D. Wooten
Quoted 12 lines · TimO
Quoted 11 lines · Ben Greear
Quoted 9 lines · Kevin D. Wooten

After reading the posts about the MODULE_LICENSE macro, I am in disbelief. I was under the impression that one could write a "closed-source" module and distribute it in binary form, and be in compliance. Please tell me I am wrong? We use Linux as a platform for some data acquisition, and we are currently distributing ( in very limited quantity to people who would already have signed an NDA ) modules that currently have no official license as yet. We are researching which license to use, but according to these post's we have almost no choice, Open Source or not at all!

Quoted 8 lines
Quoted 4 lines

No, you just can't use certain symbols if you're not GPL. If your code already works, then you're fine, as previously existing symbols will not be thus restricted... You can just make your MODULE_LICENSE == "mine-all-mine...including-all-my-bugs"

Ugghh! Don't confuse/equate MODULE_LICENSE with EXPORT_SYMBOL_GPL_ONLY; two different animals, two differnet goals. See archives for more info.

My apologies for the misinterpretation.
This scheme seems fine as long as driver related symbols are not "GPL" only.

Quoted 4 lines
Quoted 4 lines

Ben

Quoted 2 lines
(quoted diff) +0 -2
-kw
-
#6 lore raw
In reply to Kevin D. Wooten
Quoted 2 lines · TimO

Ugghh! Don't confuse/equate MODULE_LICENSE with EXPORT_SYMBOL_GPL_ONLY; two different animals, two differnet goals. See archives for more info.

What happens is a module is distributed as a combination of open-source .c files and closed-source .o files. That is, it's "mixed source" - part of the driver is open-source and part is closed-source. What happens if the open-source version of the driver is the only code that uses GPL-only symbols.
How is that handled?

#7 lore raw
Quoted 4 lines

files and closed-source .o files. That is, it's "mixed source" - part of the driver is open-source and part is closed-source. What happens if the open-source version of the driver is the only code that uses GPL-only symbols.
How is that handled?

Well then the open source bit would be GPL which would mean you can;t link it with the binary bit.

Alan

#8 lore raw
Quoted 6 lines · Alan Cox
Quoted 3 lines

files and closed-source .o files. That is, it's "mixed source" - part of the driver is open-source and part is closed-source. What happens if the open-source version of the driver is the only code that uses GPL-only symbols.
How is that handled?

Well then the open source bit would be GPL which would mean you can;t link it with the binary bit.

Ah, but what happens if I distribute the source code, the closed-source .o files, and a makefile, and tell people that they should link it? Am I violating the GPL, or is the end-user?

#9 lore raw
Quoted 3 lines

Ah, but what happens if I distribute the source code, the closed-source .o files, and a makefile, and tell people that they should link it? Am I violating the GPL, or is the end-user?

I am told by legal people you are, because you provided the code soley with the intent that it was used that way. Whether an imaginative lawyer can also get you locked away under the DMCA for distributing a device for violating copyright I dont know 8)

It is actually all very simple in legal terms. The legal terminology is "derivative work" and basically if its a derivative work of a GPL work its GPL if not it isnt. Thats what "linking" is about - not /bin/ld.

If you wanted to provide a mixed source/binary driver that wasnt derivative of the kernel (and there are lots of reasons for it) - don't GPL your open source bit use something like MPL or BSD

[I am not a laywer ...]

Alan

#10 lore raw
Quoted 7 lines · Alan Cox
Quoted 2 lines

Ah, but what happens if I distribute the source code, the closed-source .o files, and a makefile, and tell people that they should link it? Am I violating the GPL, or is the end-user?

I am told by legal people you are, because you provided the code soley with the intent that it was used that way. Whether an imaginative lawyer can also get you locked away under the DMCA for distributing a device for violating copyright I dont know 8)

But the GPL only covers distribution, not use. Just because I distribute GPL source code, doesn't mean that it actually has to work or do something. The open source portions of the driver conform to the GPL, because they #include kernel header files and, in some case, cut-and-paste from the kernel source itself. The closed source portions do not #include any kernel header files or
use any source code from anyone else.

The fact that the open source portions and the closed source portions can't function on their own is irrelevant, IMHO.

Please show me where in the GPL text it says that the act of compiling a module and loading it into memory is subject to the GPL.

Quoted 2 lines

If you wanted to provide a mixed source/binary driver that wasnt derivative
of the kernel (and there are lots of reasons for it) - don't GPL your open source bit use something like MPL or BSD

Our open source bits are GPL because they are "derived" from the kernel source, which is also GPL.

FYI, I don't consider including a header file basis for calling the module "derived" from that header file, but that's just my personal opinion and doesn't really affect this discussion.

#11 lore raw
Quoted 2 lines

Our open source bits are GPL because they are "derived" from the kernel source, which is also GPL.

Thanks I've filed that statement should it ever be useful.

Quoted 3 lines

FYI, I don't consider including a header file basis for calling the module "derived" from that header file, but that's just my personal opinion and doesn't really affect this discussion.

Interface definitions tend to be treated a little differently to "code". But as I keep trying to beat into people - if you are going to mix GPL and non GPL code see a lawyer - thats what they are there for

#12 lore raw
In reply to Timur Tabi
Quoted 5 lines · Timur Tabi

The fact that the open source portions and the closed source portions can't function on their own is irrelevant, IMHO.

Please show me where in the GPL text it says that the act of compiling a module and loading it into memory is subject to the GPL.

That'd be paragraph 2 b)

b) You must cause any work that you distribute or publish, that in whole or in part contains or is derived from the Program or any part thereof, to be licensed as a whole at no charge to all third parties under the terms of this License.

... These requirements apply to the modified work as a whole.

Since your program, which happens to consist of one open source part and one proprietary part, is partly a derived work from the kernel source (by using kernel header files and the inline functions in it) your whole work must be distributed under the GPL.

Quoted 6 lines
Quoted 3 lines

If you wanted to provide a mixed source/binary driver that wasnt derivative of the kernel (and there are lots of reasons for it) - don't GPL your open source bit use something like MPL or BSD

Our open source bits are GPL because they are "derived" from the kernel source, which is also GPL.

"open source bits" ... from "the work as a whole" ?

regards,

Rik

Signature · 3 lines
DMCA, SSSCA, W3C?  Who cares?  http://thefreeworld.net/

http://www.surriel.com/		http://distro.conectiva.com/
#13 lore raw
Quoted 5 lines · Rik van Riel

Since your program, which happens to consist of one open source part and one proprietary part, is partly a derived work from the kernel source (by using kernel header files and the inline functions in it) your whole work must be distributed under the GPL.

I contest your meaning of the word "work". The open source portion of my module is one "work", and the closed source portion is another "work". I could deliver these two works as separate tarballs, if I wanted.

Not only that, but the closed-source portion of my driver is not derived from any GPL program, so the phrase "contains or is derived from the Program" does not apply to it, because it:

1. Does not contain any part of any GPL Program
2. Is not derived from any GPL Program

Quoted 4 lines
Quoted 2 lines

Our open source bits are GPL because they are "derived" from the kernel source, which is also GPL.

"open source bits" ... from "the work as a whole" ?

The point I'm trying to make is that I can play the semantics game very easily, and still not be forced to open-source the closed-source portions of my driver. Because of the way Linux loads modules, I could take all the open-source portions and link them into one .o file, and then insmod that .o file without any problems. Then the closed-source portion would also be insmod'ed. The only issue is that closed-source portion would fail to load if the open-source portion is not already loaded.

#14 lore raw
Quoted 8 lines · Timur Tabi
Quoted 5 lines · Rik van Riel

Since your program, which happens to consist of one open source part and one proprietary part, is partly a derived work from the kernel source (by using kernel header files and the inline functions in it) your whole work must be distributed under the GPL.

Quoted 4 lines

I contest your meaning of the word "work". The open source portion of my module is one "work", and the closed source portion is another "work". I could deliver these two works as separate tarballs, if I wanted.

Let me quote you, from 2 emails back in the thread:

"The fact that the open source portions and the closed source portions
can't function on their own is irrelevant, IMHO."

The irrelevance here is IYHO ... it may well be judged that since these two portions of the work need each other in order to function, the thing really is one work.

You're well into the danger zone here...

regards,

Rik

Signature · 3 lines
DMCA, SSSCA, W3C?  Who cares?  http://thefreeworld.net/

http://www.surriel.com/		http://distro.conectiva.com/
#15 lore raw
Quoted 3 lines

The irrelevance here is IYHO ... it may well be judged that since these two portions of the work need each other in order to function, the thing really is one work.

a)
vmware for linux needs a linux kernel to work. does that meen you whant to gpl it?

b)
you can write abstraction modules for different os's. and the non-gpl module works with all of them. so my module does not need the linux abstraction module it also works with the free- BSD module. the only #ifdef in my module will be around the module registration code. or i write a propriatary loader, so that even the same binary works for different os's

Signature · 1 line
****** It's nice to be important, but it's more important to be nice ! ******
#16 lore raw
In reply to Rik van Riel
Quoted 5 lines

The irrelevance here is IYHO ... it may well be judged that since these two portions of the work need each other in order to function, the thing really is one work.

You're well into the danger zone here...

All of the licensing terms in the GPL are subject to change by
a court of law, or by legislative action. Nonethless, Copyright law is generally agreed (IMHO, and based on past discussions of this topic on this and other inappropriate forums) to be based on textual derivation rather than functional dependency.

Usual disclaimer: I am not a lawyer.

Craig Milo Rogers

#17 lore raw
In reply to Rik van Riel

Decía LLX:

Quoted 8 lines
Quoted 3 lines

The irrelevance here is IYHO ... it may well be judged that since these two portions of the work need each other in order to function, the thing really is one work.

a)
vmware for linux needs a linux kernel to work. does that meen you whant to gpl it?

Quoted 8 lines

b)
you can write abstraction modules for different os's. and the non-gpl module works with all of them. so my module does not need the linux abstraction module it also works with the free- BSD module. the only #ifdef in my module will be around the module registration code. or i write a propriatary loader, so that even the same binary works for different os's

And you can use a big array with your binary code and made the source public:
/*
This code is GPL
*/
char *code={big array of code}

void do_somenthin(){
	void (*fun) = code;
	(*fun)();
}

(well, I dosn't remember the exact sintax of pointer to funtioncs but ... )

You can put the binary driver like "microcode", and GPL

Saludos
Drizzt

Signature · 4 lines
... 10 IF "LAS RANAS"="TIENEN PELO" THEN PRINT "Windows is good".
____________________________________________________________________________
Drizzt Do'Urden              
[email protected]     
#18 lore raw
Quoted 4 lines

(well, I dosn't remember the exact sintax of pointer to funtioncs but .= =2E. )

You can put the binary driver like "microcode", and GPL

Nope. The only cases we have microcode like that in the kernel is downloaded firmware for devices. The same stuff you'd have been finding in the boot rom instead 12 months ago before the price squeezes reached ripped out any flash component and doing software download. Richard Stallman doesn't like that either, but since he currently runs an OS loaded by and using a binary only BIOS I don't have any direct sympathies right now

#19 lore raw

Yes, Alan, use these tecnique is a mess, and nobody are going to include a mess like this in a standar kernel. But someone could use this kind of tricks to include "binary only mess" in the kernel and no break the GPL ...

I also think about a dinamic patching kernel from user space from /dev/kmem, another mess but, the limits in "what is dinamic linking", is not clear in the GPL for me, and you can use these kind of tricks.

I can understand the problems with reports to linux-kernel with binary only drivers, and why is important that tainted kernel. People in this list don't have to do the work that must be done for a the company that made the driver who knows what the binary driver do.

But I can understand less the GPLONLY_ option (and I have read another mail from you where you comment that there are people that doesn't want that their code are called from "binary only modules"), because there are tricks that you can use to jump over the GPL and not put "a glue GPL module" (like I have read in the thread about this question).

I only thinking about all this topic. I prefer the source for the drivers (simple, easy debugging) , but is better a binary driver (but the support must be done by the company that make that driver,of course), that no driver at all :(, but these is only my opinion.

Saludos
Drizzt

-----Mensaje original-----
De: Alan Cox [mailto:[email protected]]
Enviado el: jueves, 08 de noviembre de 2001 13:41
Para: [email protected]
CC: LLX; [email protected]
Asunto: Re: Module Licensing?

Quoted 4 lines

(well, I dosn't remember the exact sintax of pointer to funtioncs but .= =2E. )

You can put the binary driver like "microcode", and GPL

Nope. The only cases we have microcode like that in the kernel is downloaded firmware for devices. The same stuff you'd have been finding in the boot rom instead 12 months ago before the price squeezes reached ripped out any flash component and doing software download. Richard Stallman doesn't like that either, but since he currently runs an OS loaded by and using a binary only BIOS I don't have any direct sympathies right now

#20 lore raw

RE: Module Licensing? (thinking a little more)

People always can license the binary driver with the asm source code or a asm source with:

func:

	dd 0xFF,0x01 
	...

This is a mess also :)

Saludos
Drizzt

#21 lore raw

Re: Module Licensing? (thinking a little more)

Quoted 9 lines · Drizzt Do'Urden

People always can license the binary driver with the asm source code or a asm source with:

func:

	dd 0xFF,0x01 
	...

This is a mess also :)

Read the GPL again, more carefully.

-Doug

Signature · 2 lines
Let us cross over the river, and rest under the shade of the trees.
   --T. J. Jackson, 1863
#22 lore raw

RE: Module Licensing? (thinking a little more)

Yes, clause 3.a) "machine readable source code". A .s file is, "machine readable source code" by the assembler and by people that have enough time to lost.. It is like head.S, but using numeric labels and other stuff of that kind.

Btw I don't understand exactly the problem with the use of asm code (in opcodes or in nemonics) and the GPL in this particular case . To me, it's "machine readable source code" by the assembler and if it's compilation produces exactly the same executable, and don't see the problem.

It's a nighmare to debug and mantain, but it's the problem who made the "asm" modulo not the kernel people.

Saludos
Drizzt

#23 lore raw

Re: Module Licensing? (thinking a little more)

Quoted 9 lines · Drizzt Do'Urden

Yes, clause 3.a) "machine readable source code". A .s file is, "machine readable source code" by the assembler and by people that have enough time to lost.. It is like head.S, but using numeric labels and other stuff of that kind.

Btw I don't understand exactly the problem with the use of asm code (in opcodes or in nemonics) and the GPL in this particular case . To me, it's "machine readable source code" by the assembler and if it's compilation produces exactly the same executable, and don't see the problem.

You need to read further down in section 3:

The source code for a work means the preferred form of the work for making modifications to it.

If the code was originally written in assembler, than the assembler source (with comments and meaningful variable names) is the "preferred form". If written in C and compiled to assembler, it isn't.

IANAL, but the wording of the GPL is fairly clear. Obfuscated and semi-compiled source doesn't cut it.

-Doug

Signature · 2 lines
Let us cross over the river, and rest under the shade of the trees.
   --T. J. Jackson, 1863
#24 lore raw

Re: Module Licensing? (thinking a little more)

In reply to Drizzt Do'Urden

Decía Doug McNaught:

Quoted 6 lines

You need to read further down in section 3:

The source code for a work means the preferred form of the work for making modifications to it.

Ok, thanks for the info, topic closed.

Saludos
Drizzt

Signature · 4 lines
... Insumiso se pasa el DOOM sin pegar un tiro.
____________________________________________________________________________
Drizzt Do'Urden              
[email protected]     
#25 lore raw

Re: Module Licensing? (thinking a little more)

In reply to Doug McNaught
Quoted 4 lines · Drizzt Do'Urden

Yes, clause 3.a) "machine readable source code". A .s file is, "machine readable source code" by the assembler and by people that have enough time to lost.. It is like head.S, but using numeric labels and other stuff of that kind.

Sigh.

"The source code for a work means the preferred form of the work for making modifications to it."

--
Russell King ([email protected]) The developer of ARM Linux
http://www.arm.linux.org.uk/personal/aboutme.html

#26 lore raw

RE: Module Licensing? (thinking a little more)

Quoted 3 lines

"The source code for a work means the preferred form of the work for making modifications to it."

yes, I understand all the problem with these, but you can make modifications also in "these kind of source code".

Closed theme :) I think

Saludos
Drizzt

#27 lore raw

Re: Module Licensing? (thinking a little more)

Hi.

Quoted 8 lines
Quoted 3 lines

"The source code for a work means the preferred form of the work for making modifications to it."

yes, I understand all the problem with these, but you can make modifications also in "these kind of source code".

Closed theme :) I think

Sure, but it's not your _preferred_ way of making modifications.

// Stefan

#28 lore raw
In reply to Rik van Riel
Quoted 13 lines
Quoted 5 lines

The fact that the open source portions and the closed source portions can't function on their own is irrelevant, IMHO.

Please show me where in the GPL text it says that the act of compiling a module and loading it into memory is subject to the GPL.

That'd be paragraph 2 b)

b) You must cause any work that you distribute or publish, that in whole or in part contains or is derived from the Program or any part thereof, to be licensed as a whole at no charge to all third parties under the terms of this License.

Quoted 7 lines

... These requirements apply to the modified work as a whole.

Since your program, which happens to consist of one open source part and one proprietary part, is partly a derived work from the kernel source (by using kernel header files and the inline functions in it) your whole work must be distributed under the GPL.

False. Please cite the section of the GPL version 2 that you
think means that the derived work, once created, *must be distributed*. If you read closely, I think you will find that it says that *if* you distribute the mixed work, then then it must be distributed under the the terms of the GPL. If you do *not* distribute the mixed work, then the GPL does not place any restrictions on your use of the derived work. In other words, the restrictions imposed by the GPL are conditional upon the derived work's actual distribution: no distribution, no restrictions.

Furthermore, "loading into memory" is (arguably) considered an
essential part of running a program, and section 0 of the GPL version 2 says:

The act of running the Program is not restricted,

I should insert the usual disclaimer: I am not a lawyer.

Craig Milo Rogers

#29 lore raw
In reply to Rik van Riel
Quoted 4 lines · Rik van Riel
Quoted 2 lines · Timur Tabi

The fact that the open source portions and the closed source portions can't function on their own is irrelevant, IMHO.

...

Quoted 5 lines

Since your program, which happens to consist of one open source part and one proprietary part, is partly a derived work from the kernel source (by using kernel header files and the inline functions in it) your whole work must be distributed under the GPL.

You could define a generic device driver interface. Let's call it UDI. Now you implement a GPL version of this for Linux, a 2-clause BSD version for FreeBSD, and a proprietary version for Windows. You then write a few drivers using this interface, under a half-dozen different licenses.

One of those drivers is proprietary. It can run in the Linux kernel. It sure isn't derived from Linux though. There isn't any Linux code in it, and the driver works in the Windows and FreeBSD kernels as well as in Linux.

There you go: use of GPL symbols from a non-GPL driver.

(usual disclaimer applies: not legal advice, see a lawyer)

#30 lore raw
In reply to Timur Tabi
Quoted 5 lines

Since your program, which happens to consist of one open source part and one proprietary part, is partly a derived work from the kernel source (by using kernel header files and the inline functions in it) your whole work must be distributed under the GPL.

This is obviously incorrect, that would say that

	#include <sys/types.h>

means my app is now GPLed. Good luck enforcing that.

Signature · 2 lines
---
Larry McVoy            	 lm at bitmover.com           http://www.bitmover.com/lm 
#31 lore raw
Quoted 11 lines · Larry McVoy
Quoted 5 lines

Since your program, which happens to consist of one open source part and one proprietary part, is partly a derived work from the kernel source (by using kernel header files and the inline functions in it) your whole work must be distributed under the GPL.

This is obviously incorrect, that would say that

	#include <sys/types.h>

means my app is now GPLed. Good luck enforcing that.

You're right, just including <sys/types.h> won't do that, but either of:

1) using inline functions from a .h file or
2) linking to the library/kernel later on

might mean your stuff is GPLed.

Be careful which definitions you get from the
header file, inline functions are a very grey
area ;)

regards,

Rik

Signature · 3 lines
DMCA, SSSCA, W3C?  Who cares?  http://thefreeworld.net/

http://www.surriel.com/		http://distro.conectiva.com/
#32 lore raw
Quoted 10 lines
Quoted 6 lines

This is obviously incorrect, that would say that

	#include <sys/types.h>

means my app is now GPLed. Good luck enforcing that.

You're right, just including <sys/types.h> won't do that, but either of:

Quoted 8 lines

1) using inline functions from a .h file or
2) linking to the library/kernel later on

might mean your stuff is GPLed.

Be careful which definitions you get from the
header file, inline functions are a very grey
area ;)

I think discussions of copyright issues need more precision
than "your stuff", please.

The object module resulting from compiling the code fragment
above ("#include ...") might be a derived work of both the code fragment and the (GPL'd) header file, thus subject to the restrictions of the GPLv2. The original source code itself would not be subject to the GPLv2, because "<sys/types.h>" is not itself a copyrightable string (although it could be trademarked, of course, which might lead to the same effect).

Using, in the non-GPLv2-compliant source file, any symbol name
included from the GPL'd header file, has been commonly assumed not to taint the source file itself with the restrictions of the GPLv2. This is more of a grey area, because the use of these symbols over a block of code could easily construct a derivation of a usage pattern that was copyrighted in the interface definition document.

Usual disclaimer: I am not a lawyer.

Craig Milo Rogers

#33 lore raw
In reply to Larry McVoy
Quoted 11 lines · Larry McVoy
Quoted 5 lines

Since your program, which happens to consist of one open source part and one proprietary part, is partly a derived work from the kernel source (by using kernel header files and the inline functions in it) your whole work must be distributed under the GPL.

This is obviously incorrect, that would say that

	#include <sys/types.h>

means my app is now GPLed. Good luck enforcing that.

glibc is GLPL ... not GPL ...

-dean

#34 lore raw
In reply to Larry McVoy
Quoted 11 lines · Rik van Riel

You're right, just including <sys/types.h> won't do that, but either of:

1) using inline functions from a .h file or
2) linking to the library/kernel later on

might mean your stuff is GPLed.

Be careful which definitions you get from the
header file, inline functions are a very grey
area ;)

[Please: read the whole message before flaming, OK? Thanks]

The reason they are not gray areas is this:

virus licenses such as the GPL can't cross over well defined boundaries. If they could, then the fact that the kernel is GPLed would make any application that is run on top of the kernel also GPLed. The commonly held belief that userland is somehow different than the kernel is just that, a belief. In the eyes of the law it's all a big pile of code running together. We draw the distinction between running and linking but that is arbitrary, and while it makes sense to us, I challenge someone to prove that it makes sense to the courts.

Noone believes that just running an application makes it GPLed, not even Stallman. In fact, he acknowledges the boundary issue when he says:

"These requirements apply to the modified work as a whole. If identifiable sections of that work are not derived from the Program, and can be reasonably considered independent and separate works in themselves, then this License, and its terms, do not apply to those sections..."

While some people might like you to believe that the FSF gets to interpret what "can be reasonably considered independent", they don't. Remember, just because you write something in a contract doesn't mean it is enforceable. For example, you could sign a contract that says that you belong to me and I can do whatever I want with you, but that's never going to be enforceable, for the obvious reasons. There are all sorts of laws, in pretty much all societies, which override anything too wacky put into a contract. A good example of how much this is ignored is the typical employment agreement that everyone makes you sign in California. It's not worth the paper it is written on, California has laws on the books which make that clear, yet companies use the fact that people are ignorant to get away with it. But if those same companies are challenged in court, as they frequently are, they will lose. And they know it, they are just playing the odds.

Getting back to the GPL, the point is that neither the FSF nor the copyright holders get to arbitrarily decide what is and is not separable. One could certainly try, for example, to say that running a program on GPLed kernel makes the program GPLed. It's not going to work though. Keep that in mind when thinking about what is or is not GPLed.

I think another way to look at it might be: if you extend a GPLed program using well defined interfaces, you can probably get away with not GPLing your code. You may have to fight for it and it generally isn't worth the fight, but you could win and it is likely that you'd win.

On the other hand, if you're in there changing how an existing GPLed program works, and there isn't any way to pull your stuff out cleanly, then you are definitely stuck with the GPL. And that's as it should be, you are in GPLed code. And even if you could get away with winning a court battle, you have to ask yourself if it is worth it. In my opinion, it's relatively rare that changes to the kernel are really worth the fuss.

Signature · 2 lines
---
Larry McVoy            	 lm at bitmover.com           http://www.bitmover.com/lm 
#35 lore raw
Quoted 8 lines · Larry McVoy

I think another way to look at it might be: if you extend a GPLed program using well defined interfaces, you can probably get away with not GPLing your code.

On the other hand, if you're in there changing how an existing GPLed program works, and there isn't any way to pull your stuff out cleanly, then you are definitely stuck with the GPL. And that's as it should be, you are in GPLed code.

Fully agreed. From what I read, Timur is somewhere halfway between these points and I'm not sure he can get away with not GPLing his work the way he does things now...

regards,

Rik

Signature · 3 lines
DMCA, SSSCA, W3C?  Who cares?  http://thefreeworld.net/

http://www.surriel.com/		http://distro.conectiva.com/
#36 lore raw
In reply to Larry McVoy
Quoted 3 lines · Larry McVoy

I think another way to look at it might be: if you extend a GPLed program using well defined interfaces, you can probably get away with not GPLing your code.

I imagine the question then arises to whether even the "EXPORT_SYMBOL_GPL" interfaces fall under the definition of a well-defined interface. I suppose (IANAL) that the "intent" of the author in marking this interface as GPL-only has some impact on the discussion?

Cheers, Andreas
--
Andreas Dilger
http://sourceforge.net/projects/ext2resize/
http://www-mddsp.enel.ucalgary.ca/People/adilger/

#37 lore raw
In reply to Larry McVoy
Quoted 11 lines
Quoted 5 lines

Since your program, which happens to consist of one open source part and one proprietary part, is partly a derived work from the kernel source (by using kernel header files and the inline functions in it) your whole work must be distributed under the GPL.

This is obviously incorrect, that would say that

	#include <sys/types.h>

means my app is now GPLed. Good luck enforcing that.

Your compiled object module might be a derived work, hence its
distribution would be restricted by the terms of the GPL version 2. Your source code file would not be a derived work (under certain currently widely-held assumptions about interface copyrights), and hence could be distributed without restriction by the GPL.

Usual disclaimer: I am not a lawyer.

Craig Milo Rogers

#38 lore raw
In reply to Larry McVoy

The GPL 2.0 uses language that makes it pretty clear that it was meant to describe the source code, not the object module. It uses "source code" to refer to what is licensed under the GPL several times.

The GPL has been a great source for nebulous debates without much in the way of solid ground to stand on. It's not remarkable that it has a history in academia.

Too bad those who drafted the GPL were not (competent) lawyers, either.

} Your compiled object module might be a derived work, hence its
} distribution would be restricted by the terms of the GPL version 2. } Your source code file would not be a derived work (under certain } currently widely-held assumptions about interface copyrights), and } hence could be distributed without restriction by the GPL.
}
} Usual disclaimer: I am not a lawyer.
}
} Craig Milo Rogers
} -
} To unsubscribe from this list: send the line "unsubscribe linux-kernel" in } the body of a message to [email protected]
} More majordomo info at http://vger.kernel.org/majordomo-info.html
} Please read the FAQ at http://www.tux.org/lkml/

#39 lore raw
Quoted 1 line

Too bad those who drafted the GPL were not (competent) lawyers, either.

Actually the GPL was drafted by lawyers. Any license when you dig at the edges of how to exploit it gets just as nebulous, as horribly enough does copyright law.

The merging one is analogous to other fun things like in the US if a java applet violates a patent when executed is it the web site or the browser user who broke the rules...

#40 lore raw
In reply to Cort Dougan

Alan-

Come on, this is the US. You sue BOTH the web site AND the browser user (and the browser, the browser maker, the ISP that provided access, Sun for designing Java with this capability, ...) :-)

Quoted 9 lines · Alan Cox
Quoted 1 line

Too bad those who drafted the GPL were not (competent) lawyers, either.

Actually the GPL was drafted by lawyers. Any license when you dig at the edges of how to exploit it gets just as nebulous, as horribly enough does copyright law.

The merging one is analogous to other fun things like in the US if a java applet violates a patent when executed is it the web site or the browser user who broke the rules...

Signature · 4 lines
Don Dugger
"Censeo Toto nos in Kansa esse decisse." - D. Gale
[email protected]
Ph: 303/652-0870x117
#41 lore raw
In reply to Larry McVoy
Quoted 2 lines · Cort Dougan

The GPL 2.0 uses language that makes it pretty clear that it was meant to describe the source code, not the object module.

The GPL describes _both_ the source and object code forms of the program.

See section 3, "You may copy and distribute the Program (...) in _object code or executable form_ under the terms of sections 1 and 2 above provided that you _also do one of the following:_ ..."

The terms to be followed for object code and executable form are given explicitly in section 3.

Quoted 2 lines

It uses "source code" to refer to what is licensed under the GPL several times.

Actually, section 0 defines what is licensed: "This License applies to any _program_ or other work which contains a notice placed by the copyright holder saying it may be distributed under the terms of this General Public License.

If we are picky, you are right that this does not appear to refer to an object module _if_ the object module doesn't contain the license notice. (This may perhaps be unfortunate wording in the GPL (IANAL either)). (Lucky that newer kernel modules _do_ contain a license notice, isn't it?)

But if you take that strict interpretation, you have _no_ right to copy or distribute the object module anyway, except as granted by the license on the accompanying source code.

In other words, the GPL doesn't have to "apply" to the object module to deny you the right to copy it. You start with no right to copy the object module _at all_, except as granted by the copyright owners.

-- Jamie

#42 lore raw
In reply to Larry McVoy
Quoted 3 lines · Jamie Lokier

But if you take that strict interpretation, you have _no_ right to copy or distribute the object module anyway, except as granted by the license on the accompanying source code.

Right you are, but this is trivial to circumvent if you are trying to ship a binary driver. The binary doesn't have GPLed code, you are shipping two things which are not combined when they were shipped, the end user combines them when running them.

Signature · 2 lines
---
Larry McVoy            	 lm at bitmover.com           http://www.bitmover.com/lm 
#43 lore raw
In reply to Larry McVoy
Quoted 9 lines · Larry McVoy
Quoted 3 lines · Jamie Lokier

But if you take that strict interpretation, you have _no_ right to copy or distribute the object module anyway, except as granted by the license on the accompanying source code.

Right you are, but this is trivial to circumvent if you are trying to ship a binary driver. The binary doesn't have GPLed code, you are shipping two things which are not combined when they were shipped, the end user combines them when running them.

Indeed, the very old, never resolved, "user does the link" argument again.

I honestly don't know what a court of law would decide on that. It may well be different in different jurisdictions.

-- Jamie

#44 lore raw
In reply to Kevin D. Wooten
Quoted 10 lines · Timur Tabi
Quoted 2 lines · TimO

Ugghh! Don't confuse/equate MODULE_LICENSE with EXPORT_SYMBOL_GPL_ONLY; two different animals, two differnet goals. See archives for more info.

What happens is a module is distributed as a combination of open-source .c files and closed-source .o files. That is, it's "mixed source" - part of the driver is open-source and part is closed-source. What happens if the open-source version of the driver is the only code that uses GPL-only symbols. How is that handled?

This is the approach I would use:

Kernel module:
Exports a standard interface like file or terminal (or several...) GPL all of this work.
Make it useful for others too - and you may sell some additional HW.

Propritary code (user application):
Uses standard file operations - like fopen, seek, ... The needed headers are LGPL and thus safe.

Comments?

/RogerL

Signature · 3 lines
Roger Larsson
Skellefteå
Sweden