This is one of the rare occasions where I think Theo is right

Posted Sep 4, 2007 16:01 UTC (Tue) by JoeBuck (guest, #2330)
Parent article: Relicensing: what's legal and what's right

Now, in general I'm not impressed with the BSD folks' argument that, when a proprietary software company builds on BSD code and makes the result proprietary, everything is cool, but when GPL folks take BSD code and make it GPL, this is somehow a moral violation.

But this case is different. Regardless of which license you prefer, forking the driver so that fixes cannot be shared serves no one. It would best for all if the common core of the driver can be shared by the BSDs and by Linux, especially since it originated in BSD-land in the first place. The OS-specific portions (needed since the interface to a driver is different in BSD and Linux kernels) can go into separate files and headers, and each side can use the license they prefer.

In the past, the FSF has generally used a policy of cooperation with upstream; while there have been exceptions (and I'm sure that FSF opponents will quickly point them out) people working on the GPL side of the fence have for years contributed patches back using the original license, and this is generally a good idea.

to post comments

This is one of the rare occasions where I think Theo is right

Posted Sep 4, 2007 16:06 UTC (Tue) by JoeBuck (guest, #2330) [Link] (10 responses)

I should amend my statement a bit: I don't agree with Theo's legal argument that, because the BSD notice can't be removed, it means that a given file must be dual-licensed forever. That argument would have interesting implications for, say, Microsoft; if any of their partners who've signed up for "shared source" get hold of a file that was originally BSD and later added Microsoft changes, Theo's saying that the recipient can distribute that file under BSD, even though Microsoft would strongly object. But Theo's right in that you can't strip out the license text.

I agree with Theo only in the sense that in this case allowing the two communities to share development is the right thing to do.

This is one of the rare occasions where I think Theo is right

Posted Sep 4, 2007 16:11 UTC (Tue) by cate (subscriber, #1359) [Link] (8 responses)

Copyright cannot be removed (but after expiry time), and the license is attached to copyright: it explain how the copyright-holder allows you to use the sources, so really you cannot change! You are not the owner but you decides how to use code that it is not your?

This is one of the rare occasions where I think Theo is right

Posted Sep 4, 2007 16:23 UTC (Tue) by sayler (guest, #3164) [Link] (2 responses)

So.. is this based on any actual LEGAL grounds?

This is one of the rare occasions where I think Theo is right

Posted Sep 4, 2007 16:41 UTC (Tue) by cate (subscriber, #1359) [Link] (1 responses)

IANAL ;-) but I think I read something about this in license-discuss@opensource.org . Anyway how a person who doesn't own the copyright could change a license of the file? (removing a license is like changing a license: you change how the source are redistributed). Do you see something like this in any license?

BTW the single license on a file is not necessary the same as the aggregate license of a program (source), and it not the same as the license of the generated binary program. So it is ok that your program or your binary is only GPL licensed, but no part of the common licenses allow you to relicense an existing file.

This is one of the rare occasions where I think Theo is right

Posted Sep 7, 2007 23:12 UTC (Fri) by giraffedata (guest, #1954) [Link]

Anyway how a person who doesn't own the copyright could change a license of the file? (removing a license is like changing a license: you change how the source are redistributed)

Don't confuse a copyright license with the text describing the license. Nobody can "remove" someone else's license -- it's simply not possible -- but one can remove the text describing the license from the file.

(Similarly, one can add text to a Microsoft product saying, "you may copy this as you please," but it doesn't mean a recipient of it has a license to copy it).

Removing the text does not change or eliminate the license; it just keeps people from being aware of it. The author of open source code wants all recipients to know they have his permission to copy it, hence he sets as a condition of copying the code that the copier copy it with the license text intact.

(I think in the BSD case, the author also wants to make sure every recipient of the code knows that the author disclaims liability for any mistakes he made in the code).

This is one of the rare occasions where I think Theo is right

Posted Sep 4, 2007 17:11 UTC (Tue) by elanthis (guest, #6227) [Link] (4 responses)

It's legal because the license SAID it was legal. Simple.

The license says, "you may choose to distribute this under either the BSD or the GPLv2." If you distribute it under the GPLv2, then all recipients of the work received it under the GPLv2 terms, and by those terms, they must abide by the GPLv2 (and nothing else) for all subsequent redistributions.

Note also that even though the BSD license says that it cannot be removed, that only applies if the license itself applies to the work. Since it is distributed under the GPL now, the BSD license no longer applies, and the "do not remove" clause is therefor irrelevant. There is no law against cutting up legal documents (although they are not valid if modified and not ratified by all parties involved), but no legal document was modified. The header is not by itself a legal document - it is merely a notice of which license applies to the text. Notice that the GPL itself is not in the header. The header just says, "use this license," and since the modified work is now GPL, that's all the header has to say.

That said, I still this is really lame of the Linux developers. They're being selfish. The BSD developers are likewise being lame, because they're bitching about Linux devs doing the same thing proprietary devs do. Both sides are in the wrong, ethically, in my opinion.

This is one of the rare occasions where I think Theo is right

Posted Sep 4, 2007 19:28 UTC (Tue) by proski (guest, #104) [Link] (2 responses)

That said, I still this is really lame of the Linux developers. They're being selfish.

I think you owe an apology to Linux developers. They are not selfish.

Not only wasn't the patch removing BSD license from ath5k ever applied, but the author of the patch withdrew it promptly. The idea of removing BSD license from the dual licensed files received very little support in linux-wireless list.

Incorrect patches are proposed all the time, and get shot in the review process, as it happened in this case.

This is one of the rare occasions where I think Theo is right

Posted Sep 6, 2007 7:38 UTC (Thu) by ekj (guest, #1524) [Link] (1 responses)

Sure. But there is one legitimate point coming from the BSD-camp.

Linux benefited by being allowed to copy their code.

If we then later find bugs in the same code, or make improvements to it, it would be nice of us to share those improvements and bugfixes with the BSD-people. Which means allowing those improvements to be distributed under the BSD-licence.

Currently, the code from BSD is BSD/GPL, but any improvements we add along the way, will be GPL. Which means the BSD-people aren't free to grab the improvements back.

The BSD-licence explicitly allows these kinds of things though, the major difference between GPL and BSD is precisely the fact that with GPL you need to publish your improved version under the same licence (if you distribute it anyway) while with BSD-code that is not a legal requirement.

The fact that it's not a legal requirement does however not mean it ain't the rigth thing to do.

It would be nice if anyone making bugfixes and/or improvements to this driver would explicitly licence those bugfixes/improvements under BSD+GPL, so that the BSD-people *would* be free to grab the improvements back.

I second other commenters though: If developers don't like the idea that others may take their code and then refuse to contribute improvements back, they'd be well-adviced to choose a licence that doesn't permit precisely that.

This is one of the rare occasions where I think Theo is right

Posted Sep 7, 2007 5:48 UTC (Fri) by proski (guest, #104) [Link]

My post was only about calling Linux developers "lame" and "selfish", which was, in my opinion, totally uncalled for. Such characterizations of Linux developers can only be based in their own actions, whereas your post shifts the focus towards BSD developers.

This is one of the rare occasions where I think Theo is right

Posted Sep 14, 2007 4:02 UTC (Fri) by celtic_hackr (guest, #47391) [Link]

I agree that the dual licensed code can relicensed either way. The original author has already said this.

I don't see what the issue is of people complaining about not giving back.
The code is GPL. Ergo you have the source, rewrite the measly 5% or 10% or 1% that is newly added GPL and incorporate it into the original dual BSD/GPL code and then release in BSD if you want. Come on folks are we all that unimaginative?

Where's the harm, so someone might have to do a little reading and writing.
You've got the gosh darn code! This is a copyright matter not a patent matter people, you can rewrite!

Granted, I think the Linux guys should have just left the copyright alone. They're all, both sides, acting like spoiled children. Heck my toddler has better manners. But then, I raise her that way.

This is one of the rare occasions where I think Theo is right

Posted Sep 4, 2007 19:41 UTC (Tue) by nim-nim (subscriber, #34454) [Link]

In theory, you are right, sharing development would be the right thing.

In this particular case I doubt it would happen. There is a lot of bad blood between the OpenBSD and the Linux wireless communities. (public harassment of Intel for doing *Linux* wireless by Theo, campaign against OLPC because the wireless stuff was not being opened fast enough, shameless procrastinating when OpenBSD was caught misappropriating Linux GPL code, etc). OpenBSD wireless advocacy is lynch first, think later, never present excuses.

Theo only claimed brotherhood once it become clear he had no legal leg to stand on. There can be no doubt his preference would have been to publicly immolate his "brothers". Pretty suicidal to bully when your licensing model makes you rely on third-party goodwill if you want my opinion. Especially now the Linux wireless people seem to have settled on a wireless stack at last and should start pumping drivers.

One of the first feedbacks on the proposed patch was to play nice with OpenBSD. However that was before the public smear campaign I doubt anyone feels ready to tell the Linux wireless devs how to license their code now.

This is one of the rare occasions where I think Theo is right

Posted Sep 4, 2007 16:50 UTC (Tue) by epa (subscriber, #39769) [Link] (4 responses)

I think Theo's point is that the spirit of the BSD licence is more important than the letter. Yes, it doesn't use legal force to compel you to share your code back, but the spirit of BSD suggests that you should do so. Interestingly, RMS has said the same thing: the exact legal meaning of the GPL is less important than its spirit, which is that you should share code with others and not impose restrictions on users.

The position that dual-licensed files must 'always' remain dual-licensed after making changes seems a bit untenable though. If that were the case, then any dual-licensed BSD+GPL code incorporated into OpenBSD would have to remain dual-licensed forever as well. This would arguably mean that proprietary vendors would no longer be able to strip out the GPL and include the code in proprietary software. But I fear we are getting into legal nonsense land here.

The commonly accepted meaning of dual licensing is a disjunction. You can distribute this under licence A, or licence B, at your option. (Or continue distributing under A+B.) It's unfortunate that the OpenBSD people think it means something else. To avoid misunderstanding we may need a new term to mean dual-licensed-choose-either-yes-really-I-mean-it.

This is one of the rare occasions where I think Theo is right

Posted Sep 4, 2007 20:25 UTC (Tue) by sepreece (guest, #19270) [Link] (2 responses)

"The position that dual-licensed files must 'always' remain dual-licensed after making changes seems a bit untenable though. "

I think the reality is more like "The original file and the code it contains always remain dual-licensed." Once you modify the file, the resulting file is licensed under a combination of the original dual-licensing and whatever license you put on your own modifications. Depending on your choice, this could even make the modified file undistributable, even by you (if your license terms aren't compatible with either of the original licenses).

Personally, I would be happiest ethically with retaining the original dual licensing for any modifications. Anything else just seems rude.

Then what's the point of dual licensing?

Posted Sep 5, 2007 19:02 UTC (Wed) by mrshiny (guest, #4266) [Link]

If it's unethical to release changes to a dual-licensed file using only one license, what's the point of HAVING a dual-license in the first place? Seems to me the original author is saying "Take your pick", not creating some super-license that contains all the terms of the individual licenses. And IANAL but as far as I'm concerned if a file is available under two licenses you can use the privileges of one license to strip away the other license. The original code would still be dual-licensed but your modified version would be under whichever license you chose.

This is one of the rare occasions where I think Theo is right

Posted Sep 6, 2007 9:50 UTC (Thu) by jschrod (subscriber, #1646) [Link]

But the original author stated his intent that it is choose and pick. (And the FreeBSD folks did so, too.) And in law, intent counts a lot.

That said, I also think that Linux changes to the drivers should be licensed BSD+GPL, in respect for the original authors' work. It's not a legal thing, it's a moral obligation, IMHO.

This is one of the rare occasions where I think Theo is right

Posted Sep 4, 2007 21:35 UTC (Tue) by xtifr (guest, #143) [Link]

> I think Theo's point is that the spirit of the BSD licence is more important than the letter.

Then how come they don't complain when people release binaries with improvements that will never benefit the BSD community? Frankly, I love BSD (both the license and the system); I grew up in Berkeley and have many friends in the BSD community, but the rabid anti-GPL fanaticism of <em>some</em> members of the BSD community has always set my teeth on edge. I mean, even if you <em>do</em> buy the theory that GPL software is not truly free, how is making source code available for inspection (if nothing else) worse than not doing so?