Free Software Licenses Revocable?

There is nothing stopping you from GPLing something, and then using your code in a proprietary system. Free software licenses grant rights to someone who doesn't have them in the first place. They do _not_ waive copyright, because then they couldn't be enforced.

Absent a revocation clause in a license, if you get a copy of some free software, the author can't take away your rights to modify and redistribute it. They can change the license on future versions, but you can continue to develop the old version. (Look at the openssh history.)

You should also take a closer look at the UCC and how it applies to software... the shrinkwrap licenses which come with commercial software are a completely different beast and generally notify you that you are entitled to nothing more than the physical media of your purchase, and that you should be happy if you can use the software in any way.

He goes on, "if you get a copy of some free software, the author can't take away your rights to modify and redistribute it". Now, this is the topic under discussion. If Fred doesn't know (and it's his job to know things like this), what makes Alan so sure?

Microsoft bought a lot of IP apparatus from SGI; did they get the copyrights to the code SGI released under GPL, and can they take the rights back? Can somebody buy out Trolltech AB and screw over the whole KDE project, and Apple's new browser besides? Can CMU decide to add a special exception to the Mach license to squeeze Apple for its use in Macosix? Can BSD make a special exception and squeeze MS over their TCP/IP stack? Does the answer to questions like this depend on where the license appears, in the file or on a web site?

The answers to the above questions seem pretty important to the Free Software movement. If it's really not clear, we had better modify the licenses to make it clear.

"If Fred doesn't know (and it's his job to know things like this), what makes Alan so sure?"

"The answers to the above questions seem pretty important to the Free Software movement. If it's really not clear, we had better modify the licenses to make it clear."

Frankly, if such thing were possible, don't you think that someone would have raise the issue from the start?

Free Software permit forking. For instance Savannah is a fork from SourceForge now proprietary. Do you think that VA Software, now a proprietary software company, can forbid Savannah, which is based on GPL version of SourceForge?

If so, whole Free Software idea is broken. But licensing itself is meaningless.

Frankly, I think you are completely wrong. I'm not a lawyer but I'm pretty sure that once you get a license, nobody can remove the rights the license give to you until you respect the license. If not, each license, not especially Free Software licenses, is a lie, since licenses we talk about does not mention a timeline (if so, the problem is different, but we speak a Free Software).

Long answer: In general, software licenses are governed by contract law. The details will vary from one jurisdiction to another, but it's usually not so easy to simply revoke a contract you have made with another party simply because you no longer wish to honor the terms of the contract.

AlanShutko was correct when he said that free software licenses do not in general waive any rights. The original author retains full copyright of a work released under, say, the GNU GPL. They simply offer a nonexclusive license to parties who agree to certain terms. This is no different from proprietary software, except that you generally obtain a license for proprietary software by paying money to the copyright holder, rather than simply agreeing to a set of conditions. Whether you have paid money for your license or obtained it otherwise, it cannot be taken from you simply because someone holds the copyright to the licensed code.

[Aside: I have read that the GPL does lose standing under contract law in some jurisdictions because the "contract" does not involve a mutual exchange between both parties, the way a purchased license does. I don't know enough about contract law to comment on this.]

By the way, the Free Software foundation explicitly addresses this question in the GPL FAQ:

Can the developer of a program who distributed it under the GPL later license it to another party for exclusive use?

No, because the public already has the right to use the program under the GPL, and this right cannot be withdrawn.

The Creative Commons Licenses (which are not intended for software, but are similar to many Free Software licenses) are not revocable. This is not stated explicitly in the license text, but it is mentioned on the baseline rights page. The irrevocability is a consequence of the licenses' lack of any sort of termination or expiration.

Copyright law has nothing whatever to do with contract law. There is no contract involved. Anybody who talks about copyright licenses in terms of contracts is fundamentally confused.

A copyright license is just a public statement of conditions under which the author pledges not to exercise the legal monopoly powers afforded to the author under copyright law.

Alan, Matt, the FSF, and the Creative Commons can assert irrevocability until they turn blue, but in the absence of written law and case law it's just wishful thinking. Wishful thinking is a poor substitute for sound legal analysis. As far as anybody reporting here has found so far, that analysis has not been done. I'm not satisfied with wishful thinking, and I'm frankly disappointed in anybody who is.

The whole problem appears to go away for licenses that assert irrevocability explicitly.

The FSF's view is an opinion, but so is ncm's: is the promise that a copyright license makes binding on the person making that promise? Eben Moglen says yes, and ncm says this is wishful thinking. As a professor of copyright law, Moglen's opinion might be credited by most as worth more than ncm's, but let's not go there for now.

But ncm claims that the problem goes away if the license asserts irrevocability explicitly. But why does this make a difference? He's already claimed that the copyright license is in no sense a contract, and can be cancelled any time by the author. But if the person making a promise can break his word, why does it matter if he says "I really, really mean it"?. If the GPL can be revoked, I don't see why a court can't order someone to revoke a license that makes a stronger promise (say, because the work infringes a patent or someone otherwise brings pressure on the copyright holder, or even because the author is hard up for money and wants to kill off competition from his own previous free software release).

You also may be missing the point that if the FSF makes good arguments that pretty much everyone buys, and the arguments don't seem particularly weak, then decades might go by before anyone challenges those arguments in court, so they have the practical force of law.

I don't think it's so easy to state that copyright licenses are not contracts. Rod Dixon (a lawyer and law professor) has written several analyses stating his opinion that the GPL and similar licenses are contracts and rely on both copyright law and contract law. On the other hand, the FSF's legal counsel Eben Moglen has written that "Licenses are not contracts."

Moglen argues that the GPL is not a contract because it does not rely on voluntary assent by the licensee; rather it grants them certain rights which they would not even have in the absense of the license, and any "restrictions" placed by the license are simply rights which the licensor has chosen not to give away. Dixon on the other hand states that violations of free software licenses are not necessarily also violations of the licensor's copyright, and in these cases the violation of the license must be treated as a breach of contract. In these cases, the reasoning in ProCD v. Zeidenberg (deciding that shrinkwrap licenses are contracts) would seem to apply equally to free software licenses.

I'll concede that the GPL may not be a contract, but I don't think that this is obvious, or that I am "fundamentally confused" for examining it in the context of contract law.

See http://cr.yp.to/softwarelaw.html for details. In most jurisdictions you can't retroactively retract a waiver. And if you could, what would a waiver be good for, anyway?

ncm wrote:

A copyright license is just a public statement of conditions under which the author pledges not to exercise the legal monopoly powers afforded to the author under copyright law.

I don't have an opinion on whether copyright licenses are implicitly perpetual. I don't know, and have not found anyone in a position to say who does. Fred von Lohmann ought to be qualified to say, but he says it's not clear.

If I were a judge, and had no precedents to fall back on, I might note that the author could have said "perpetual" in the license but chose not to. I am worried that if somebody did retract a license, a judge might embrace such an argument. If I were managing a distribution, I would probably have to obey a retraction, because I couldn't afford to go to court with enough legal help to establish a good precedent, and would have no real confidence of winning in any case.

Suppose a hunters' cabin had a sign up on it for years that said, "Come on in, just clean up after yourself." Then one day you arrived and it said "No trespassing". Would you expect a judge to rule that the owner had given up the right to control his property because the original sign didn't have a time limit?

Why would the word "perpetual" make a difference, and not just as well be revoked along with the rest of the license? Without it, the author, like the cabin owner, has promised not to enforce his privileges for the time being. After giving reasonable notice of the change, he can then reassert his privileges without contradiction. He can't prosecute people for publishing his work up to that point (that would be a retroactive change) but he can warn that henceforth he will prosecute.

With the word "perpetual", the author is promising not to ever prosecute his privilege. It would be a stretch for a judge to accept an argument that the author hadn't really meant it at the time. People are allowed to bind their future actions, but you can't assume it. They have to say so. If the copyright is sold, the buyer knows the license says "perpetual", so ought to have equal difficulty rescinding it.

I can't imagine why Matt Brubeck insists the X copyright license conditions are contractual. Is it just because the language in them resembles language in contracts? It doesn't have any language in it identifying actions that imply agreement by the reader. If the license said something like, "By exercising the rights we have extended above, you agree to the following," then the following terms would seem to form something like a shrinkwrap contract. Most Free Software licenses steer wide of contract law, apparently because copyright law is pretty simple compared to it. Contract law is riddled with centuries' worth of holes bored through and through it by clever lawyers.

I don't think contract law would help resolve the question in any case. In most jurisdictions, a contract may be broken at will by either party -- unless it says otherwise in the contract.

Dan Bernstein's page citied by Fefe seems to mix up fair use and UCC rights. That doesn't detract from its intended use, but it doesn't say anything helpful to this issue.

For the record, I would like to find that the GPL license is implicitly perpetual. It would save a great deal of work, and risk. I just don't trust wishful thinking enough to believe it without some hefty citations. Adding "perpetual" to the license seems like very cost-effective insurance.

I am not a lawyer, ohgodohgodohgodoeeeeeeyaaaaaah...(gurgle)

For countries that based their codelaw on ancient roman one, everything not explicitely forbidden is legal.

It means that a license is "perpetual" until the license mention a timeline.

The only case where revoking a license is possible is when someone disrespect some of the license's clause. If there no time clause, the debate end.

Sure, I'm not a lawyer. Anyway, I have some knowledge in ancient roman's law and I still have some logic.

Dixon writes: Instead, many mass market software licenses are intended to impose restrictions on users of software that have nothing to do with copyright. Since these restrictions are outside of the scope of copyright, software developers must rely upon another body of law (other than copyright) to enforce these restrictions, if a user violates the license; that body of law is contract law.

Law degree or no, I think Dixon is missing the point (but I'm not lawyer, so my opinion should be taken with a grain of salt). The GPL imposes no restrictions beyond those in copyright law, and the original author receives no consideration (money, etc.) from users. The GPL is a waiver of certain copyright restrictions under certain conditions. So I don't really see how you could consider the GPL to be a contract.

As for the issue of revocation: I don't know what the rules are for revoking copyright waivers, if any such rules exist in any given country. But I don't think this is quite as big a problem as it initially appears--most significant free software projects are a collection of files with multiple copyright holders, and the different copyright holders may be able to assert the GPL against each other.

If this were the case, a copyright holder who revoked their GPL license might be blocked by the other contributors' GPL licenses from distributing the software under a different license.

Blocking distribution might be just what the revoker has in mind. That's what I'm worried about.

I agree that the GPL is essentially a "pure copyright" license (though in addition to a non-exclusive copyright license, it also contains promises in sections 4 and 6 that the licensor will offer identical terms to future recipients of the program).

So, laying aside contract law, let's look at the GPL in terms of copyright law only. In the US, termination of copyright licenses is governed by Section 203 of the Copyright Code, which allows for termination 35 to 40 years after a license is granted. In Rano v. Sipa Press the Ninth Circuit Court of Appeals ruled that because of section 203, nonexclusive copyright licenses of unspecified duration cannot be terminated before this 35 year date, despite California state law allowing contracts to be terminated at will unless otherwise specified.

However, other courts have vehemently disagreed with Rano v. Sipa, according to a this NY Law Journal article. The article shares ncm's view that copyright licenses with unspecified durations may or may not be terminable at will. The authors recommend (as ncm does) that terms should be made explicit, given the uncertain precedent in this area.

[Interestingly, it is state contract law that has allowed copyright holders to revoke licenses in these cases. If the license were governed by US copyright alone, as the court held in Rano v. Sipa, then it could be terminated only under very specific circumstances.]

I was also incorrect in my earlier post when I said that the Creative Commons licenses did not specify a duration. In fact, they explicitly grant a perpetual license, so it seems that Lessig has a solid basis for claiming that the CC licenses are irrevocable.

In summary, ncm's concerns are valid. Courts have ruled both ways on this issue. Making the license term explicit is the safest course.

I am not a lawyer.

Two problems have been identified. The first is that, if the owner hasn't specified a duration for the license, a judge might well (and with plenty of justification) consider it revocable at any time.

The second problem is that without a contract, the copyright owner might not be bound at all. He might just change his mind and prosecute his privilege anyhow (perpetual or not, retroactive or not), and a judge might agree that this is within his rights under copyright law.

The first problem is easy to fix, particularly for programs that use the GPL as recommended, where the user may choose any later ("perpetual") version of the GPL.

The second problem is trickier. What does it take to bind the copyright owner? A contract. What does a contract need? Acts of agreement by both parties and exchange of "consideration". In the absence of money, what might serve as consideration? Code, of course. The first time a copyright owner accepts a patch under the license, consideration has been exchanged, and both parties have acted on the license terms, forming an unwritten contract.

The significance of accepting the first patch is something I have not seen mentioned anywhere before.

I exchanged private mail with gmp, a lawyer, who suggests that licensees abiding by the license terms might itself serve as consideration -- publishing modified source, in the case of the GPL. For the BSD and X licenses, it's hard to identify any consideration; they only say what not to do. That might make the BSD and X licenses more readily revocable than the GPL (presuming first that they have all been made explicitly perpetual; without that, a contract wouldn't seem to have any obligations).

I think we have a problem. It might not be that bad, but I think we need a GPL version 2.1. Here's why.

So far, when people have been talking about the enforceability of free software licenses, they have been looking from the perspective of the author. In this respect, Eben Moglen and Richard Stallman are correct. If you are a licensee of a piece of free software that is under the GPL, you can't possibly have _more_ rights than are set forth in that license, because copyright law by default reserves all rights for the author. Therefore, your only source of rights is the license, and you must be bound to its terms. It doesn't matter whether it is a contract or not.

The problem is that nobody seems to have looked from the perspective of the user/licensee. Where does the licensee of a piece of free software get rights to continue to use that software? If a free software license were not a contract, then I can see nothing that binds the author to its terms. As ncm noticed, if nothing binds the author to the terms of the license, then nothing prevents the author from revoking that license. Think of it this way: a license is a promise not to sue for infringement (not exactly accurate, but a reasonable approximation). You can stop promising whenever you want, unless something binds you otherwise. In practice, this is done by having your lawyer draft a letter that says "your license to use such-and-such is hereby terminated as of <date>. Signed _____" Just imagine IBM getting one from alan.

I am of the opinion that a free software license does form a binding contract, from the moment the licensee does something that is beyond what copyright law would permit them to do in the absence of a license (though there is plenty of uncertainty as to where to draw that line, and plenty of disagreement among the licenses that are currently in use). In contracts law terms, distributing a license with your software constitutes an offer, and the user's action in doing something otherwise prohibited by copyright law, constitutes an acceptance. For that to be binding, there needs to be "consideration" in the contract -- that is, the author needs to get something in return for the license granted. I think this is not a problem for the GPL, which imposes restrictions on the licensee that benefit the author (enforced publication of source code for distributed modified binaries). On the other hand, BSD users may be SOL.

Just consider this: whether or not you believe a free software license is a contract, we have a problem. If it is a contract, then it does not specify a termination date, and therefore may be terminable at will by either party. If you believe it is not a contract, then you are in the position of arguing that it is somehow more binding on the licensor than if it were. I think that is a losing argument.

Most state law seems to say that a contract is terminable at the will of either party if it does not specify a duration, or the terms by which it may be terminated. The GPL does not do that. However, a fair amount of case law holds that the state law rules on this issue do not apply to copyright licenses, and that licenses are assumed to be for the duration of the copyright term if they do not specify otherwise. See, e.g. Nimmer on Copyright.

"where an assignment or license does not expressly prescribe the period
or term of its duration, it will generally be construed (in the absence
of evidence of a contrary intent) to be effective for the duration of
the then existing copyright term of the work." 3 Melville B. Nimmer &
David Nimmer, Nimmer on Copyright § 10.10[F], at 10-98 (1998)

See also Viacom v. Tandem, 368 F.Supp. 1264 (S.D.N.Y. 1974). I am not sure whether this reasoning also holds for "gratuitous" (non-contractual) licenses.

A few more recent cases have seriously clouded the issue. There is a line of cases on the issue of whether 17 U.S.C. 203 of the Copyright act preempts state contract law on the issue of termination of licenses. The first is Rano v. Sipa (9th Circuit, 1993), which mbrubeck described above.

More recently, the 7th Circuit court in Illinois, in the case of Walthal v. Rusk (1999), came to the opposite conclusion, holding that 17 USC 203 sets a maximum term, not a minimum, and that state law is therefore not preempted. Illinois contracts law, like California law, holds that licenses are terminable at will if they do not specify a term. A more recent case in the 11th Circuit called Korman v. HBC followed Walthal. I think the weight of legal academic opinion is strongly against the Rano decision and in favor of Walthal and Korman. However, as it stands there is a split in the circuits which the Supreme Court may have to resolve. These laws and precedents were made to protect starving artists from predatory record companies. They make absolutely no sense in the free software context. But they are what we've got. From a practical standpoint, free software authors may be free to revoke their licenses, depending on where they (or their licensees) live.

Another case at the district court level in New York, TV Globo v. Brazil Up-Date Weekly (1999), held that copyright licenses that do not specify a duration are effective for the remainder of the copyright term of the work. This case is interesting because under New York law contracts of unspecified duration are terminable at will. The judge in TV Globo held that a different rule applies for copyright licenses than for contracts, and cites tons of precedent for that view. So far I don't have a good idea of how to reconcile TV Globo with Walthal.

A simple answer to this problem is for the GPL to include a license grant of the form commonly used in intellectual property agreements. Something like what the Open Software License has:

1) Grant of Copyright License. Licensor hereby grants You a world-wide,
royalty-free, non-exclusive, perpetual, non-sublicenseable license to
do the following:
...

Some references:

The Walthal v. Rusk decision can be found here: http://ngro_obsrvr.tripod.com/articles/decision.html

The 11th Circuit decision is Korman v. HBC, which can be found here: http://www.law.emory.edu/11circuit/aug99/97-5064.man.html

By the way, while I am an associate at a law firm, a law school graduate, and I have passed the bar exam, I am not yet licensed. Even if I was, of course, this would not be legal advice.

Here's what Lessig has to say on his licenses (which have explicit language making them "perpetual"):

on the permanence of cc licenses

... CC licenses are, at this moment, at least, permanent, in the sense that the term is as long as copyright runs [...] But just because you can't revoke a particular license doesn't mean you can't revoke the offer. If, for example, you offer content under a CC license for a month, and then change your mind, you can stop offering the content under that license. Anyone who accepted your offer while it was valid, of course, has a deal. But no one after you withdraw the offer can accept anymore.

From the point of view of a free software/open source licensee, I don't know if it's possible for a licensor to revoke their licence, but I do know that if they do, I want consideration. The courts are likely to grant it to me, too, if my business depends on your software.

As I said previously, with no explicit duration limitation, normally there should be no limitation in time, as mbrubeck confirm ("in fact, they explicitly grant a perpetual license").

But anyway, the next ncm's quotation speaks also for me: yes, "you can stop offering the content under that license" and yes, "no one after you withdraw the offer can accept anymore" from you. But the first one who have "a deal" can have "a deal" with other people.

I download gcc and I have a deal with gcc authors: I can read, modify, execute, redistribute the code. It's my deal, and it's irrevocable according to Lessig. Whatever the fact the gcc team can stop making such deals with other people, I still can read, modifify, execute and... redistribute... and deal with other people.

As I said previously, with no explicit duration limitation, normally there should be no limitation in time.

I download gcc and I have a deal with gcc authors: I can read, modify, execute, redistribute the code. It's my deal, and it's irrevocable according to Lessig.

... CC licenses are, at this moment, at least, permanent, in the sense that the term is as long as copyright runs [...] But just because you can't revoke a particular license doesn't mean you can't revoke the offer. If, for example, you offer content under a CC license for a month, and then change your mind, you can stop offering the content under that license. Anyone who accepted your offer while it was valid, of course, has a deal. But no one after you withdraw the offer can accept anymore.

(Once again yeupou substitutes wishful thinking for law, and even for reading carefully. That works fine as long as nothing you do matters, but in that case picking your nose works just as well.)

Interestingly, the case of Gcc is unusual. Every substantial bit of the Gcc distribution has had a copyright assignment to the FSF, so it is not possible for any author of any code in Gcc to withdraw it. Lots of people complain about the copyright assignment policy, and lots of great code (not least Mark Mitchell's great work while at SGI, IIRC, and a lot of Intel optimization work) has been excluded for that reason. Now it looks like the FSF was right and the complainers were wrong.

One strategy that I think I heard from Linus, and that I am applying to XPLC is exactly the reverse of that of the FSF: I try to have as many copyright owners as possible, with as much fragmentation as possible.

By doing this, I am getting insurance against me going insane. No individual is in power of reverting the Linux kernel to a different license, not even Linus himself. It would be simply impossible. You remember the trouble that Mozilla.org went through when they changed the license on Mozilla? They were missing the agreement of a few people, which was blocking the whole process (these persons could simply not be found!).

I agree this is a risk, but one that I am willing to take. Also, there is the fact that only a (or "the"?) copyright owner(s?) can protect their work, this is one of the reasons that the FSF invokes.

Let's be generous and say that there's a one percent chance, each, that you or any of your contributors will go insane. If everybody assigns their rights to you personally, then there's a one percent chance that you will get fed up and take it away from everybody. Without assignments, if you have have 20 contributors, you all suffer a (1 - (0.99)^20) = 18 percent chance that one will withdraw his contribution. With 40, you have a one in three chance of the software being held hostage.

Now look what happens if everybody assigns to the FSF. The FSF can only withdraw the software if its board votes that way. I don't know how many are on its board; I will guess seven. If it takes four crazy votes, that's a likelihood of about 0.01^4, or 1 chance in 100 million of the software suddenly disappearing.

What this means is that your "insurance against going insane" is itself insane. With each added contributor you multiply the risk of your code being held hostage.

Anyway, can you give me a pointer to gnu.org that confirm that copyright assignation to the FSF has something to do with our discussion too in the FSF point-of-view?

According to http://www.gnu.org/evaluation/evaluation.html it's about who will enforce the GPL in case of violation.

Finally, your answer is off-topic regarding to my post.

Please note that everybody that officially assigns copyright to the FSF gets a signed copy of a contract with the FSF that says something like:

FSF agrees that all distribution of the Works, or of any work "based on the Works", or the Program as enhanced by the Works, that takes place under the control of FSF or its agents or successors, shall be on terms that explicitly and perpetually permit anyone [...], to redistribute copies of the work to anyone on the same terms [...]. These terms shall not restrict which members of the public copies may be distributed to [...]. These terms shall not require a member of the public to pay any royalty [... etc, etc, etc ...] offered in the form of machine-readable source code [... yada, yada, yada ...]

So the FSF may not have to keep distributing such a program under the original license (persumable GPL), but it will have to be some sort of Free Software copyright notice. This has been done so that you don't have to worry about the FSF board of directors going crazy and suddenly turning the program into something proprietary. But it still allows changing the license a bit when necessary. I would only assign the copyright to a program to someone else if it would come with similar guarantees that the program would always be distributed as Free Software.

Well, I might be crazy to think that my scheme will work, I haven't thought all that much about it, but I've read this from an interview with Linus, and frankly, even if Linus was to pack his beans and go home, that wouldn't change much to the code, and he'd have a pretty hard time proving what's his and what isn't.

I wrote: "Can somebody buy out Trolltech AB and screw over the whole KDE project, and Apple's new browser besides?"

Private communication from someone at Apple reveals that their Safari browser does not use Qt at all. Hence, the answer to my question, for Apple, is a firm "no". KDE as a whole remains in danger, albeit little more so than much of the rest of the Free Software world. At least the official GNU projects that have had copyright assignments are safe.

This doesn't mean that Apple is safe, over all. Lots of the code they use is from the NetBSD and FreeBSD contributors, who might, in principle, decide to squeeze Apple. That seems unlikely, since any contribution by an unstable individual can probably be replaced cheaply.

ncm writes, with great certainty, Anybody who talks about copyright licenses in terms of contracts is fundamentally confused. He also claims not to be a lawyer, so from whence comes his confidence in his understanding of contract law? (My guess is "from his butt").
-russ