LWN: Comments on "FSF responds to Oracle v. Google and the threat of software patents"

What Oracle Java stuff is GPLv2 anyway?

BenHutchings — Thu, 16 Sep 2010 19:24:17 +0000

A Java runtime environment under GPLv2 is useless to Google, because they do not want to require applications to be licenced under GPLv2. The key difference in licencing is that the free J2SE has a 'runtime exception' while the free J2ME does not.

Trackback

coriordan — Mon, 13 Sep 2010 23:18:10 +0000

One of your worse blog entries :-/

It's a barrage of errors, exaggerations, and strange interpretations. There are just so many mistakes I'd like to correct, I don't know which ones to bother with...

Trackback

FlorianMueller — Mon, 13 Sep 2010 11:48:00 +0000

I've now also done a blog posting on the FSF statement, and in that one I make reference to this discussion here.

FSF responds to Oracle v. Google and the threat of software patents

drag — Sun, 12 Sep 2010 18:26:36 +0000

yes. They are horrible and only a very large company can hope to absorb them and stay healthy.

The overhead of maintaining a patent system like this is far more costly then to have no software patents in the first place.

Unfortunately the FSF thinks the GPL end justifies any misleading means

boog — Sun, 12 Sep 2010 15:39:25 +0000

And on the final part, you quoted me talking about the "licensor" (which in this scenario would be Oracle) and then talk about "licence they chose" (which would be Google), so you misunderstood something.

Please. I meant Sun (not Google) chose to release Java under the GPL. Oracle can't undo that choice now and their hands are tied by whatever is promised by that licence, which they didn't write. It might have committed them to an extensive patent grant. To what extent that is the case may require a lot of legal argument to sort out, but Oracle cannot retrospectively reinterpret the licence according to their present intentions. (Of course, as copyright holders they do have the right to make further releases under whatever licence they choose.)

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sun, 12 Sep 2010 15:16:08 +0000

Regarding your arguments on estoppel etc, those appear to be principles based upon general behaviour and public statements

I just explained what the legal theories for an implicit patent license are under US law.

Those principles are not just about general behavior and public statements.

I would be very surprised if the specific language of the licence that Sun/Oracle willingly entered into had no bearing.

I didn't say it has no bearing, but that a patent holder can argue that the scope of a copyright license is one thing and the scope of a patent license another. The theories I mentioned would indeed take the license into account, but even the broadest possible copyright license doesn't have to result in a broad patent license. RMS himself emphasizes all the time how very different those rights are (which is why he rejects the term intellectual property).

Unfortunately the FSF thinks the GPL end justifies any misleading means

boog — Sun, 12 Sep 2010 14:30:38 +0000

We're obviously going to have to agree to disagree.

Regarding your arguments on estoppel etc, those appear to be principles based upon general behaviour and public statements. I would be very surprised if the specific language of the licence that Sun/Oracle willingly entered into had no bearing.

However, a licensor might want to grant a very broad copyright license and reserve a lot of rights concerning his patents.

Indeed they might want to, but the whole point is that their hands are tied by the licence they chose. Interpretation is out of their hands now.

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sun, 12 Sep 2010 13:51:48 +0000

Thanks for highlighting the quotes. You're persistent, but I'll answer again, though I hope it will be the last time for today :-)

I don't understand why an implicit licence "must be reasonable", with "reasonable" being defined without regard to the wording of the licence. I think that is a completely illogical step. Such a principle could only be imposed by an external law, but you haven't cited one that might be relevant.

There's no such thing as an "Implicit Patent License Act" or similar law that I could present and that would deal in detail with this. Depending on jurisdiction, there would be different legal theories. Since Oracle-Google is a US case, the theories that could lead to an implicit patent license under US law are called legal estoppel, equitable estoppel, conduct, and acquiescence. To discuss the differences between the four would be a major digression, but especially the one that contains the word "equitable" makes it clear that some fairness/reasonableness principles comes into play.

I agree with you that if there's a copyright license that grants extremely far-reaching rights to licensees (and yes, the GPL does go quite far), one might want to argue the implicit patent license should also go very far. However, a licensor might want to grant a very broad copyright license and reserve a lot of rights concerning his patents.

In order to attack a GPL Dalvik, Oracle could find itself in the tricky position of having to show both that the GPL Dalvik infringes AND that the GPL Java from which it would have been derived makes NO use of the patent (beacause otherwise the implicit licence would operate).

Oracle would only show that the hypothetical GPL Dalvik infringes. Google would then, in that hypothetical scenario, argue that GPL Java (such as OpenJDK) also does, and that Google just used it on GPL terms. Then the question would be whether it's a new use of the relevant patent(s); and we discussed before what that means.

And it might be possible to copy verbatim the specific code from the GPL Java, surely making it impossible to argue that the patent uses were different.

I explained before that there are various kinds of changes that could be considered a new use. If the patents are really at the heart of the virtual machine, even changes to seemingly separate parts of the overall program could be seen as a new use. Just imagine if Google had added another command, and to execute any command, everything has to go through some patented code...

I tend to believe the FSF people who have given most thought to their licence; they haven't often been wrong regarding the legal strength of the GPL where it has been tested. Certainly, I think Oracle would have a vastly more difficult case if they were attacking a GPL Dalvik. So, no, I don't think the FSF position is misleading.

I didn't base anything I said on any statistics in terms of how often the FSF had been right. What was quite apparent in its statement was a desire to promote the GPL as particularly patent-safe, or at least to give that impression to many readers of the statement. In my opinion, the GPLv2 has its strengths that are beyond doubt, but protection against patent attacks isn't one of them, and worse than that, since redistribution of modified code is a software freedom, it's pretty bad in my view that the FSF promotes something as a "strong defense" that is the weaker the more someone exercises that freedom (freedom 3 according to the Free Software Definition).

I explained before how I view this, and I'd like to leave it at that.

Unfortunately the FSF thinks the GPL end justifies any misleading means

boog — Sun, 12 Sep 2010 13:16:31 +0000

...could you please improve the layout of your comments

I've tried.

If it's an implicit one, then there's no language of a grant to refer to and the scope will be what a court would consider a reasonable scope of an implicit license.

Theoretically, an explicit license could be anything from extremely narrow to extremely broad, while an implicit one will be resaonable.

I don't understand why an implicit licence "must be reasonable", with "reasonable" being defined without regard to the wording of the licence. I think that is a completely illogical step. Such a principle could only be imposed by an external law, but you haven't cited one that might be relevant.

Imagine a licence, not so different to the GPL, that says "you hereby grant irrevocably any and all rights you possess that could in any way be used by you or third parties to impose additional restrictions on the receipient's use, distribution or modification of the program." That would be an implied patent grant. And it would be broad. I see no reason why it should be interpreted "reasonably" because it is implicit.

That's exactly my point. In order to attack a GPL Dalvik, Oracle could find itself in the tricky position of having to show both that the GPL Dalvik infringes AND that the GPL Java from which it would have been derived makes NO use of the patent (beacause otherwise the implicit licence would operate). I think that would be significantly more difficult than simply establishing infringement, especially as most of their patents ARE related to Java. And it might be possible to copy verbatim the specific code from the GPL Java, surely making it impossible to argue that the patent uses were different.

Of course, they would first try to argue that there is no or only a limited patent licence implied in the GPL2, but we've covered that discussion above; it's NOT established jurisprudence.

To return to the very original issue. I tend to believe the FSF people who have given most thought to their licence; they haven't often been wrong regarding the legal strength of the GPL where it has been tested. Certainly, I think Oracle would have a vastly more difficult case if they were attacking a GPL Dalvik. So, no, I don't think the FSF position is misleading.

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sun, 12 Sep 2010 11:03:32 +0000

First a request, could you please improve the layout of your comments by highlighting quoted passages more clearly? I see some people doing it with color, while I use the blockquote tag, but the way you quote entire passages in quotes (which passages then sometimes also contain quotes) is a bit confusing. Thanks.

I agree that there are two questions - is there a patent licence and, if there is one, what is its extent.

I don't deny that those are important questions, but the extent of the license has a lot to do with whether it's an implicit or explicit one. If it's an explicit one, then the language of the explicit grant defines it. If it's an implicit one, then there's no language of a grant to refer to and the scope will be what a court would consider a reasonable scope of an implicit license.

if it is effective and broad, who cares if it is implicit ? :-)

As I just explained: the scope depends on the nature. Theoretically, an explicit license could be anything from extremely narrow to extremely broad, while an implicit one will be resaonable, which is neither of the two extremes.

And I simply don't buy the argument that any change in the whole of Java means that none of the patents are being used as they were in the Sun/Oracle version

The most conservative position on the reasonable scope of an implicit license would be that any change can put the program outside that scope. That's similar to a programmer pointing out that any change in the program can make any other part unstable. It doesn't mean that it necessarily happens, but if all you know is that something changed, then that's the risk to assume (meaning you'd have to test everything again).

If a patent holder wants to limit the scope of the implicit patent license as much as possible, he will argue that whatever changes were made turned the program into something new that wasn't intended to be allowed under the license. Looking at the Oracle-Google example, the patents Oracle asserts appear to relate to the overall functioning of the virtual machine as opposed to individual programming commands. In the event Google changed anything about the commands (supporting fewer or supporting additional ones, or a combination of both), Oracle could argue that the selection of programming commands going through the patented mechanisms wasn't meant to be allowed under the implicit license.

The phrase says "limited to the use" not "limited to the code".

That's true but not helpful in practical terms. Assuming there is an implicit patent license in place and you take the relevant code and just change variable names, the algorithm itself remains the same. So it's the same use and would also in my understanding still be licensed. But once you change the logic of the relevant code, the patent holder will claim it's a new use. Since the implicit patent license should be reasonable, the nature and stature of any changes to the logic of the code should be taken into account. Absolutely minor changes should not deprive the forker and his users of the patent license, but once the changes are significant, the patent holder may have a case that the implicit license really just relied to what he (the patent holder) published.

Sun/Oracle would then have to argue that patents (originally taken out to protect Java) are used DIFFERENTLY in their Java and in the (GPL) Dalvik.

That's right. As I mentioned further above, even if there's simply a difference in terms of supported commands or whatever else that gets processed by the patented algorithms, Oracle could argue that it's a new use. If I say "could argue", that doesn't mean they'd be certain to win, but I said earlier on in this thread that the way the FSF wanted its statement to be understood was that the GPL would have solved the problem, so any reasonable doubt is enough to say that the FSF overstated.

("even if you receive the code direct from the owner of the patents" was referring to the little discussion of the "REdistribute" wording in the GPL2, where it seemed to me that the protection was less clear if you received the code direct from an owner with patents.)

Thanks for clarifying. The term "protection" is difficult to use here. If someone takes a GPL'd work and contrary to the REdistribute part simply imposes restrictions, then you are the licensee are still not directly protected. However, the forker who did that would have distributed code under the GPL without permission to do so. So in the end both of you would have a problem. The forker would have a copyright problem but you might still have a patent problem :-)

Unfortunately the FSF thinks the GPL end justifies any misleading means

boog — Sun, 12 Sep 2010 10:37:21 +0000

(On the issue of whether there is a patent licence and what its limits might be.) I agree that there are two questions - is there a patent licence and, if there is one, what is its extent. I was just arguing that if a court decides there is a licence, the uncertainty about whether there was one is resolved and that uncertainty has little bearing on the extent of the licence, which must then be decided and can be very broad. You seem to argue that because there is uncertainty about the existence of the licence it must therefore be limited. I don't think that is logically correct.

"You're just confusing people with your own confusion now. The European Commission clearly said that the GPLv2's patent grant is implicit. Even the passage you quoted says "this implicit license"."

And the passage also says "essentially received an irrevocable patent-licence"; if it is effective and broad, who cares if it is implicit ? :-) (See beginning of this comment regarding separation of existence and extent questions.)

"It's not just a limitation to the patents whose teachings are practiced in the relevant code. Note that the passage you quoted also says "would be limited to the use that is being made of the patent claim by the code as originally released under the GPL." Once any of that code is modified, let alone if entirely new code uses the same patents, you're no longer talking about the original use."

The phraseology seems ambiguous - it can (should) be read as meaning that only the use of the patent must be unaltered for protection to be effective, not that the code itself must be unaltered. The phrase says "limited to the use" not "limited to the code". And I simply don't buy the argument that any change in the whole of Java means that none of the patents are being used as they were in the Sun/Oracle version. (Such a strict limitation is certainly not intended in the implicit patent licence.) Surely the analysis would come down to specific code portions and whether the algorithmic approach is the same in original Java as in (GPL) Dalvik. Sun/Oracle would then have to argue that patents (originally taken out to protect Java) are used DIFFERENTLY in their Java and in the (GPL) Dalvik. That would be especially difficult if the code portion was the same in the two (remember this is all about a hypothetical GPL Dalvik that would be recreated by derivation from GPL Java), offering a significant potential for working around infringement claims.

("even if you receive the code direct from the owner of the patents" was referring to the little discussion of the "REdistribute" wording in the GPL2, where it seemed to me that the protection was less clear if you received the code direct from an owner with patents.)

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sun, 12 Sep 2010 08:58:02 +0000

'm not sure you're being self-consistent here. Last comment around you were implying that there must be a reasonable limit on a patent grant, this time you are saying that there isn't a patent grant.

Not inconsistent, but once those statements get taken out of context, they may appear to be. The sentence you quoted about Sections 6 and 7 of the GPLv2 not constituting a license grant meant that they don't constitute an explicit license grant (the context in which I said so). The GPLv2 as a whole (of which Sections 6 and 7 are just parts) does constitute an implicit one. When I said there must be reasonable limits, I referred to the implicit grant. When I said Sections 6 and 7 don't constitute a grant, I meant they don't constitute an explicit grant. In my opinion they don't even constitute an implicit one just by themselves (it's the GPLv2 as a whole).

As I understand it, some patent grants need not specifically list affected patents

Right, but that's not the issue. Listing patents is one thing; saying specifically "I, the right holder, grant you a license" is another.

For what it is worth (I don't have a great opinion of European bureaucracy), this lawyer seems convinced that there would be a strong patent licence granted via the GPL even if you receive the code direct from the owner of the patents.

You're just confusing people with your own confusion now. The European Commission clearly said that the GPLv2's patent grant is implicit. Even the passage you quoted says "this implicit license".

I don't understand why you say "even if you receive the code direct[ly]". The "even" is unclear to me. Why would indirect receipt be better?

Yes, the laywer does say that such protection _may_ only be limited to the patents exercised in the released code (though no reason for this limitation is given), but in our case, protection against all Sun/Oracle patents as used in Java would still be a _huge_ win.

It's not just a limitation to the patents whose teachings are practiced in the relevant code. Note that the passage you quoted also says "would be limited to the use that is being made of the patent claim by the code as originally released under the GPL." Once any of that code is modified, let alone if entirely new code uses the same patents, you're no longer talking about the original use.

Unfortunately the FSF thinks the GPL end justifies any misleading means

boog — Sun, 12 Sep 2010 08:44:16 +0000

"The key problem with Sections 6 and 7 of the GPLv2 is that even though you claim they're broad, they don't constitute a license grant."

I'm not sure you're being self-consistent here. Last comment around you were implying that there must be a reasonable limit on a patent grant, this time you are saying that there isn't a patent grant. As I understand it, some patent grants need not specifically list affected patents. (For instance, the famous covenant not to sue from Microsoft certainly did not specify the infringing patents, and I'm sure there a other examples of more direct grants or licences where all patents are not listed - how does cross-licensing usually work?)

"... about the limitations of an implicit grant I cited Dan Ravisher as well as the European Commission."

Well, I read the zdnet article citing Ravicher you cite. Possibly confused, as cioriordan says, so maybe we shouldn't put too much weight on it. I think Ravicher is rather addressing the concern of companies that licence to _all_ of their patents is potentially granted through the "right to modify " part of the licence. So the uncertainty could lead to too strong a licence rather than too weak a licence in his opinion; your description of "legal uncertainty" is therefore not the whole story.

The quote from the European Commission lawyer is rather positive, I find (http://en.swpat.org/wiki/GPLv2_and_patents ).

"Suppose, for example, that the owner of MySQL has a patent with a claim Z covering a software feature and MySQL implements this software feature and releases the implementation under the GPL. Any licensee would indeed, as the notifying party has argued repeatedly during the proceedings, essentially receive an irrevocable patent-license comprised within the GPL. However, this implicit license would be limited to the use that is being made of the patent claim by the code as originally released under the GPL. If the licensee now changes the code in a way that adds another use or implementation of claim Z it may be liable for patent infringement as regards the code it has added to what it had originally received under the GPL. "

For what it is worth (I don't have a great opinion of European bureaucracy), this lawyer seems convinced that there would be a strong patent licence granted via the GPL even if you receive the code direct from the owner of the patents. Yes, the laywer does say that such protection _may_ only be limited to the patents exercised in the released code (though no reason for this limitation is given), but in our case, protection against all Sun/Oracle patents as used in Java would still be a _huge_ win. After all, Sun must have focused its efforts on protecting Java itself.

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sun, 12 Sep 2010 04:44:10 +0000

Some of this has previously been raised by Ciarán (cioriordan) and we exchanged views. I don't want to have too much duplicity here now, but let me respond to parts of this.

The key problem with Sections 6 and 7 of the GPLv2 is that even though you claim they're broad, they don't constitute a license grant. Those sections can be fully complied with by an author who simply doesn't hold any patents on the code he publishes on GPLv2 terms. In fact, that is the way I interpret those sections: they reflect a desire that no one should publish patent-encumbered stuff under the GPLv2 (nor to do deals with third parties that result in patent encumbrance). But if someone does anyway, that text isn't an explicit license grant; one can only argue that one is implicitly needed, and about the limitations of an implicit grant I cited Dan Ravisher as well as the European Commission.

We could speculate a lot about why that clause hasn't been tested in court. So far I'm not aware of any case in which a patent holder published something under any FOSS license (GPL or whatever) and later took users or forkers to court over it. The real problem is always third-party patents. The FSF dreads the notion, and I don't like the status quo either, but the fact of the matter is that a number of major companies with all of the resources and knowledge in place to understand patent issues have decided to agree to pay patent royalties on Linux and probably also on some other GPL'd software.

Finally, your last point about "redistribute" in Section 6: you are right that the first sentence only relates to a situation where someone has already received a license from the original publisher and then passes the program on to someone else, possibly with modifications. So in Oracle/Sun's case, it's indeed hard to see how Section 6 would apply. Even if the other sentences of that section did (a kind of aggressive interpretation given that they are in the REdistribution context), there would still be the problem I outlined: the whole GPLv2 thrust is about not being patent-encumbered as opposed to granting a perpetual, irrecovable, worldwide royalty-free license.

FSF responds to Oracle v. Google and the threat of software patents

paulj — Sun, 12 Sep 2010 04:16:29 +0000

I'd just like to add my "-1, Wrong" to the parent comment, to back up coriordan. There is no "field of use: restricted to mobiles" with OpenJDK. The parent is just confused. The situation is:

The OpenJDK is under the GPL, full stop.
Additionally, the OpenJDK has a "classpath exception" clause granting permission to non-GPL Java apps to link with the OpenJDK class library.

That's all that matters in the context of this article.

There are further matters like the patent grant Sun gave to any fully-conforming Java implementation, or the fact that Sun opensourced some additional JavaME specific libraries without the classpath exception (thus many JavaME distributors chose instead to get a commercial licence from Sun/Oracle). However, those are irrelevant here as they do not apply to Google - Google's implementation is not Java and they're not using either the extra code or the APIs from JavaME in Android. These matters are just red-herrings to the discussion.

Unfortunately the FSF thinks the GPL end justifies any misleading means

boog — Sat, 11 Sep 2010 19:44:13 +0000

The relevant section appears to be (http://en.swpat.org/wiki/GPLv2_and_patents )

"6. Each time you redistribute the Program (or any work based on the Program), the recipient automatically receives a license from the original licensor to copy, distribute or modify the Program subject to these terms and conditions. You may not impose any further restrictions on the recipients' exercise of the rights granted herein. You are not responsible for enforcing compliance by third parties to this License."

(Section 7. does not seem relevant, because Sun/Oracle is not defendant in the law suit).

So section 6 implies extremely broad licence "...copy, distribute or modify the Program", even broader than I suggested. I don't see why it should be interpreted narrowly. I'm sure it was the intention of the drafters of the GPL to prevent patent attacks on free software as much as they were able. Note that even modifications are expressly allowed and no restriction of those rights may be imposed.

You say "there must be a reasonable limit" and "the implicit grant would really go too far". But I don't see how you can base those views on the language of the licence or the intention of the drafter. The licence could quite plausibly have stated "You and your successors in interest hereby agree not to bring patent suits against any GPL software", which would have been even further-ranging but perfectly explicit.

The weakness of the section seems to me to be that it has not been tested in court (because nobody has been optimistic of success?). It may be that it will be overridden by other laws (which could be where your "reasonable limit" comes into play), but you haven't been very precise about which laws they might be. Reasonableness as defined by Mueller may not be generally binding.

NB. I wonder whether Sun/Oracle "REdistribute" Java, which is the word appearing in section 6.

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sat, 11 Sep 2010 19:04:03 +0000

Section 6: "...You may not impose any further restrictions on the recipients' exercise of the rights granted herein. ..."
I don't see how patents would be excluded from that.

It's a broad statement and not imposing restrictions is not the same as granting a license. By not owning any relevant patents, one can also satisfy the requirement you quoted.

Comncerning Section 7, it seems that patent license deals are done by companies using GPLv2 software. If they only use the software themselves (like Amazon and Salesforce), that's no problem, but the likes of HTC redistribute. I'm just describing what's happening; if it were up to me, there wouldn't be any patents of that kind in the first place. So by just looking non-judgmentally at what's going on in the market, it looks like Section 7 is either easily circumvented or toothless...

Unfortunately the FSF thinks the GPL end justifies any misleading means

coriordan — Sat, 11 Sep 2010 18:56:08 +0000

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

I don't see how patents would be excluded from that.

Section 7 might not be useful in the Oracle v. Google case, since there are no third party patents, but it can be useful in other cases since it implies that if you rely on a patent licence, and if you have the authority to pass that licence on to recipients, then you do so. (Otherwise you'd be violating copyright.)

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sat, 11 Sep 2010 18:20:50 +0000

Those sections don't contain a patent license grant. Section 6 doesn't even mention patents and Section 7 says that distribution of a program under the GPL isn't allowed if there's a problem with third-party patents that can't be resolved in a GPL-compliant way. It's more like the whole thing was written with the idea in mind that someone publishing code under the GPL shouldn't hold patents on such code.

Unfortunately the FSF thinks the GPL end justifies any misleading means

coriordan — Sat, 11 Sep 2010 18:08:19 +0000

I think the GPLv2's implicit patent grant is specific to the GPL.

It's in sections 6 and 7.

If someone wanted to argue that a BSD licence had an implicit patent grant, then that would be harder to justify. There's nothing at all written, so it really would be down to implications intended by the general act of publishing something as free software. But with GPLv2, the "implicit" grant is written in the licence.

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sat, 11 Sep 2010 17:47:12 +0000

Surely the point is that any uses of code covered by patents held by its distributor are protected by the (implicit) GPL2 patent grant. In this particular case, *any* Sun/Oracle patents reading on *anything* done by GPL Java are granted to *all* derivative code of same. What is not granted is licence to patents that are not exercised in the original code.

You're being too optimistic.

There must be a reasonable limit somewhere. A minor modification of the relevant code might not be sufficient to lose the benefit of the implicit license grant, but a fundamental architectural change probably would.

Without a limit, the implicit grant would really go to far. The question isn't if there is one. It's where it is.

Unfortunately the FSF thinks the GPL end justifies any misleading means

boog — Sat, 11 Sep 2010 17:32:11 +0000

(Florian Mueller said) "But the implicit patent grant under the GPLv2 does not help if modified code infringes the same patent."

I don't think this is right. Surely the point is that any uses of code covered by patents held by its distributor are protected by the (implicit) GPL2 patent grant. In this particular case, *any* Sun/Oracle patents reading on *anything* done by GPL Java are granted to *all* derivative code of same. What is not granted is licence to patents that are not exercised in the original code.

In other words, "The implicit patent grant under the GPLv2 should be of great help if derivative code infringes on the same patent."

As I mentioned in a comment above, it seems to me that it would be quite feasible for Google to recreate Dalvik as a derivative work of GPL Java. It would be very interesting if Google did this:
- there would be a really free telephone
- the GPL2 patent grant might be tested in court (I suspect it has remained untested for good a reason)

where we see things differently

coriordan — Sat, 11 Sep 2010 16:42:16 +0000

I can't think of any cases where we could find legal interpretations of "new use". If I see one, I'll document it.

where we see things differently

FlorianMueller — Sat, 11 Sep 2010 16:10:51 +0000

This can be clarified.

You see FSF's message as being "Google should chosen GPL for the code they wrote". I see their message as being "Google should have built on IcedTea and should not have been afraid of GPL".

I didn't mean the code Google wrote. Of course the FSF argued on the basis of IcedTea. My criticism relates to the way the FSF statement conflates issues and misleads people. I previously said you highlighted Google's circumvention of copyleft in a perfectly appropriate way. But the FSF put this into the context of the GPL's ability to preserve freedom. That's misleading, and in my view it's even cynical because a patent grant that doesn't (or at least doesn't reliably) allow modifications is certainly not a way to protect software freedom.

[..] European Commission's statement. I read their statement as saying that the grant applies the "the use that is being made of the patent claim by the code as originally released", so you can change the code, and your still covered for that use of that claim of that patent.

In my opinion, if a segment of code that requires a patent license is modified, that is a new use of the patent. It's not a totally new one, sure; but it's a new one and the patent holder can argue that such new use requires a new patent license.

where we see things differently

coriordan — Sat, 11 Sep 2010 15:49:56 +0000

I think there are two core differences in how we see the situation.

You see FSF's message as being "Google should chosen GPL for the code they wrote". I see their message as being "Google should have built on IcedTea and should not have been afraid of GPL".
(I'm guessing here) We're focussing on different words in the European Commission's statement. You see the European Commission's statement about an implied grant existing only for "the code as originally released" as meaning that modifying the code cancels the grant. I read their statement as saying that the grant applies the "the use that is being made of the patent claim by the code as originally released", so you can change the code, and your still covered for that use of that claim of that patent.

Of course, I think I'm right :-)

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sat, 11 Sep 2010 15:40:38 +0000

Using the GPL isn't the point, the point is that Google should not have let their fears of the GPL prevent them from using IcedTea.

If they had said it the way you just said it, I wouldn't have called it misleading. You properly separate the GPL-or-not (copyleft-or-not) question from the question of which license does a better job as a protective shield with respect to patents. But the FSF phrased it in a way that touted the GPL as a "strong defense".

I, too, believe that Google wanted to avoid copyleft, but they might also simply have liked that Apache-licensed Harmony code for technical reasons, and regardless of the fact that they forked, they had a goal in mind (Dalvik) that apparently wouldn't have been compatible with the license grant contained in the Java specs and, I believe, wouldn't have been achievable without modifying Oracle/Sun's code to an extent that the implicit patent grant wouldn't have survived (or that it's at least doubtful it would have survived, which is enough to argue against a "strong defense" claim).

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sat, 11 Sep 2010 15:35:06 +0000

I agree it's not a GPL problem. It's an author's rights problem that would affect any license, including proprietary closed-source licenses, just like a GPLv2-style implicit patent grant isn't a unique strength of GPLv2 either.

The only reason I mentioned those limitations is to point out that the goal of granting to licensees all of the rights of the original licensor generally isn't achievable, with respect to different types of IPRs.

Unfortunately the FSF thinks the GPL end justifies any misleading means

coriordan — Sat, 11 Sep 2010 15:33:25 +0000

> conditioning it just on Google's choice of a license

I think they were conditioning it on Google building on IcedTea. Using the GPL isn't the point, the point is that Google should not have let their fears of the GPL prevent them from using IcedTea.

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sat, 11 Sep 2010 15:31:23 +0000

Thanks. Please note two things about what the European Commission decision on Oracle-Sun says in this context:

1) They defined relational database management systems as a worldwide market for purposes of competition law, so they were also interested in the situation under US law, although it's true that the Oracle case for now is just a US case and in that respect different. But if the Commission had seen a difference between those major jurisdictions, it would have pointed it out. Otherwise all of its statements referred to what was defined as the relevant geographic market in the decision.

2) The Commission's view of the implicit grant was that there was doubt about forks. While the Commission's decision made reference to the scenario of patents being infringed only by newly added code, that was just an example of where the implicit grant would face limits; they didn't say this was the *only* concern. They provided an example on which pretty much everyone can agree, just like you did. But the problem of code changes relating to existing features is a serious one I believe.

Unfortunately the FSF thinks the GPL end justifies any misleading means

coriordan — Sat, 11 Sep 2010 15:30:19 +0000

> inalienable author's rights

But that's a copyright problem, not a GPL problem. The argument you were making is that the GPL isn't as good as FSF says it is. Inalienable author's rights have never been found to be a problem for free software, and if they are a problem then it's not FSF's fault, and in fact, GPL actually does more to protect software from this problem than most (or all) other free software licences.

The GPL's requirement to maintain a changelog has the role of protecting the reputation of the author, which might reduce the risk that an author could try to use those inalienable author's rights to place restrictions on modified versions.

Unfortunately the FSF thinks the GPL end justifies any misleading means

coriordan — Sat, 11 Sep 2010 15:22:27 +0000

FSF said that the implied patent grant works in the USA but that the problem was that it mightn't work in other countries. (Eben doesn't usually speak for FSF but in that 2006 presentation he was presenting an FSF project.)

I think I remember RMS giving the UK as an example of where it mightn't work.

The Oracle v. Google case is in the USA, but the European Commission's comment is interesting. They did confirm that an implied grant existed, even if they put narrower limits on it than we'd like.

It seems clear to me that if you take OpenJDK (with features A, B, C) and you add more code (features X, Y, Z), then all downstream versions (including through an intermediary such as IcedTea) have a patent grant for features A, B, and C. And that's in the narrow interpretation of the EU.

Maybe it'd be possible to build something like Dalvik while staying within those limits - I don't know, it would depend on how many of the changes are additions, how many are changes to the existing code, and whether those changes are "substantial" in the eyes of a judge. If the USA has a broader view of implied patent grants, then it's more likely that Google would have been safe in the USA.

(I'm surprised by Dan Ravicher's comments. I think he's said elsewhere that the implied grant does work. I wonder if the second-hand reporting of his statements changed them slightly.)

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sat, 11 Sep 2010 14:34:58 +0000

It's fairly well known that the intent of the GPL is to convey all rights you, as a (primary) distributor have.

There are clear legal limits to that. For an example, some jurisdictions have inalienable author's rights. And regardless of jurisdiction, only the original author can decide to grant others a non-GPL license to the same code. The GPL tries to go very far, but it faces limits even in connection with copyright, and even more so in connection with patents.

I say "misleading" because the GPLv2 tries to solve the problem, it just hasn't been tested in the field of patent rights yet.

In connecion with this item and others you raise, let's please bear in mind that the FSF describes the GPL as a protective shield that "could've provided a strong defense against Oracle's attacks." It says "could've" (conditioning it just on Google's choice of a license); it doesn't even say "might have". My position is that if the FSF claims the GPL would have provided a "strong defense", I don't have to prove that the FSF's interpretation of the GPLv2 in connection with patents is definitely wrong to disprove their claim of a "strong defense". In this case, if there's serious doubt, which Dan Ravicher himself admitted back in 2004 and the European Commission found earlier this year, that's already enough to call the "strong defense" claim misleading.

I say "misleading" because the GPLv2 tries to solve the problem, it just hasn't been tested in the field of patent rights yet.

Just like the previous item, the thing to consider is that the FSF uses a wording that talks about a "strong defense" and while you believe that "could've" means the same as "might have", the problem with "could've" is that contrary to your claim it does not contain a concession of "could not have"; instead, it suggests to readers that Google "could have" opted for that "strong defense".

And even in the "does not help" case, the original program can be modified to your heart's content outside of the areas that implement the patented techniques. So if Oracle was asserting patents on code in areas that google wasn't planning to modify anyway, and these would have come verbatim from IcedTea, they would have a license for those patents

What you say is that if Google's virtual machine had been either IcedTea or a Google fork of IcedTea where any modifications versus Oracle/Sun's original code are in other areas than the code segments that (according to Oracle) infringe on those patents, they'd be fine.

The language contained in GPLv3 tries to achieve that situation, and whether that would be enforceable remains to be seen.

GPLv2, however, only contains an implicit patent grant. There's reasonable doubt that the implicit grant survives any modification whatsoever.

Apart from that, knowing what Google built with Dalvik, it's so different from Oracle/Sun's own Java stuff that I can't see how Google could have delivered that kind of technology without code changes in all of the critical areas (and a code change in only one area can be enough to result in a patent problem).

That's another premise of the FSF statement that I criticized in my original comment: the anti-free reasoning that software shouldn't be modified or even fundamentally rewritten, or that a standard shouldn't be implemented in a modified or even fundamentally different form.

So even in the worst case you painted (without calling attention to that), there is a fair chance that Google might have benefitted at this point from having derived Dalvik from a GPLed implemantation of Java.

All disagreements between us aside, this statement of yours shows yet another aspect of how the FSF's statement misleads people (and I raised that one in my original comment). The FSF wants people to think that the GPL has a unique trait that would have helped. But none of what you or the FSF said really is GPL-specific. It just so happens that Oracle/Sun released certain Java-related code under the GPL. If Oracle/Sun had done so under the Apache license, or any other major FOSS license, Google wouldn't have been in any weaker position than under GPLv2.

The FSF makes a connection between the superstrong superfree GPL and the subject of patent defense when the GPLv2 is actually weaker from a patent point of view than, for instance, the Apache license.

One can try -- like you do -- to analyze each individual part of the FSF's claims and then try to defend each one separately. But my comments are based on the impression the FSF seeks to generate. In my opinion, the FSF should not stoop to cheap propaganda methods. There's nothing here to say in favor of GPLv2 that has anything to do with the design of the license.

I stand by my assertion that you use your visibility in the anti-software patent arena to piggyback a decidedly anti-copyleft agenda, your "I'm pro-GPL." rhetoric not withstanding.

Actually I don't even see the "copyleft" angle here at all. What we're discussing here in connection with patents has nothing to do with copyleft. It's purely a question of whether the GPLv2 has any unique trait in connection with patents that would have helped Google. Sure, if Google had forked some GPL'd code, it would be bound to copyleft, and I actually pointed out in my original comment that Android isn't a copyleft system. One can criticize Google for not having made Android a copyleft thing without having to support utterly misleading claims that the GPL is a protective shield in connection with patents. The non-copyleft Apache license 2.0 is better for patent purposes than the GPLv2. In my opinion, it's just logical to focus on the patent part when the FSF makes a patent claim. No one doubts that the GPLv2 serves the copyleft purpose. That's not the issue.

I don't pursue an anti-copyleft agenda in this context, nor in any other. If you've ever seen any statement of mine that appeared to be part of an anti-copyleft agenda, I'll gladly take a look at it and explain.

Unfortunately the FSF thinks the GPL end justifies any misleading means

Zack — Sat, 11 Sep 2010 13:59:55 +0000

>Concerning the two items you commented with "citation needed", what citations did the FSF provide? None.

from the statement:
"The GPL is designed to protect everyone's freedomfrom each individual user up to the largest corporationsand it could've provided a strong defense against Oracle's attacks. "

It's fairly well known that the intent of the GPL is to convey all rights you, as a (primary) distributor have. That those rights stretch over various fields governed by different "intellectual property" laws is implicit, and allows for various new schemes someone might think of in the future. Just because the GPL hasn't been used so as a deciding factor in the conveying of patent using rights, doesn't mean it's unclear on the subject, even if the GPLv3 is even more explicit about it.

So your assertion that "If the GPLv2 solved the [patent] problem, the FSF wouldn't have changed its approach to patents so fundamentally with GPLv3."
is misleading. I say "misleading" because the GPLv2 tries to solve the problem, it just hasn't been tested in the field of patent rights yet. And the GPLv3 hasn't changed fundamentally at all. Its primary directive is, like the GPLv2, the safeguarding of the 4 freedoms, against which software patents are a threat.

So when the FSF states that "the GPL [...] could've provided a strong defense against Oracle's attacks", it's not "utterly misleading" because the protection is in the license, even if it has never been tested before.
Moreover, the statements says "could have" which implies "could not have", but it would have been a better position to start from Google than the position they are in now regardless.

>The other would be the idea that Oracle/Sun granted a patent license through publication of certain program code under the GPL. I also explained that in detail for you and everyone else to verify.

It's not the details you mentioned. It's the details you omitted.

"There's no way Google could impress Oracle by claiming it built Dalvik (had it done so) as a fork of IcedTea. IcedTea is a Red Hat fork of code Oracle/Sun made available under the GPLv2. But the implicit patent grant under the GPLv2 does not help if modified code infringes the same patent."

That's a "GPLv2 might not help", not a "GPLv2 does not help". By your own words there's a "degree of legal uncertainty", remember ?
And even in the "does not help" case, the original program can be modified
to your heart's content outside of the areas that implement the patented techniques. So if Oracle was asserting patents on code in areas that google wasn't planning to modify anyway, and these would have come verbatim from IcedTea, they would have a license for those patents.

So depending on the legal environment, the patents being asserted, and the actual modifications made, Google could have better defense had they based their implementation on IcedTea; they would be in a more preferable situation than they are now.
This directly contradict what you have been saying. So even in the worst case you painted (without calling attention to that), there is a fair chance that Google might have benefitted at this point from having derived Dalvik from a GPLed implemantation of Java.

>The European Commission decision related to the scenario of MySQL forks. Since MySQL is available under the GPLv2 and the same applies to whatever Oracle/Sun published under the GPL in connection with Java, the parallels are pretty clear.

And even in that hypothetical untested worst case where a patent license is only applicable to a single instance in the as-is form as it appears in the code as first distributed, it is still preferable.

I stand by my assertion that you use your visibility in the anti-software patent arena to piggyback a decidedly anti-copyleft agenda, your "I'm pro-GPL." rhetoric not withstanding.

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sat, 11 Sep 2010 13:05:30 +0000

Actually I didn't find Eben Moglen's name anywhere in the FSF statement on Oracle vs. Google. So this is unrelated.

Concerning the two items you commented with "citation needed", what citations did the FSF provide? None.

I explained, however, in detail that there are only two possible theories Google could try to hold against Oracle. One would be the license grant that comes with the Java specs (and I linked to Bruce Perens's blog, where you can read clearly that it is a grant that has nothing to do with the GPL and that actually prohibits subsets or supersets). The other would be the idea that Oracle/Sun granted a patent license through publication of certain program code under the GPL. I also explained that in detail for you and everyone else to verify.

The European Commission decision related to the scenario of MySQL forks. Since MySQL is available under the GPLv2 and the same applies to whatever Oracle/Sun published under the GPL in connection with Java, the parallels are pretty clear.

Again, let me assure you it's not a matter of "beef with the FSF and specificially Eben Moglen". The latter is not even mentioned in that analysis. As for the FSF, I do agree with them on some of what they say, but their attempt to promote the GPL as a protective shield under which Google would be safe needs to be debunked. That would also be the case if anyone else of a similar stature had made such a statement. The world is littered with entities I contradicted on such issues.

I forgot to point out that I don't have any doubts about RMS's integrity. He might just need more sincere lawyers in some situations.

Unfortunately the FSF thinks the GPL end justifies any misleading means

Zack — Sat, 11 Sep 2010 12:54:44 +0000

[...]

>The problem is that the moment IcedTea departs from the official Java specs, Oracle could use patents against it,cand the GPL wouldn't effectively prevent it from doing so if the code Oracle/Sun released under GPLv2 was modified.

citation needed

>That fact was also recognized by the European Commission in its decision on the Oracle-Sun merger.

again, citation needed, preferrably something independent from own article on your own blog giving your own view on the European Commision's decision, which actually says very little to support this "fact", specifically in relation to the case had IcedTea been used as a basis for deriving Dalvik.

I'm not sure what your beef with the FSF and specifically Eben Moglen is, but it shines through so prominently in your writing that I find it hard to read it as unbiased information anymore, right up to the point where I feel the need to double check citations where given to see if they're not out of context or otherwise contorted to make sure they're not "utterly misleading" themselves.

FSF responds to Oracle v. Google and the threat of software patents

Lennie — Sat, 11 Sep 2010 10:18:41 +0000

These taxes are however very unpredictable. Especially for smaller companies this could mean then of their business.

Unfortunately the FSF thinks the GPL end justifies any misleading means

FlorianMueller — Sat, 11 Sep 2010 09:48:14 +0000

Not least because it worked so well for MySQL's business, I'm pro-GPL. However, there are many other good FOSS licenses out there, and it depends on the purpose which one is best.

It's disappointing to see the FSF promote its GPL with statements that are utterly misleading (a diplomatic substitute for a three-letter word) such as "[t]the GPL [...] could've provided a strong defense against Oracle's attacks."

Others have already pointed out the key issues, but I'll sum up all of the issues. I have, as some of you know, spent a lot of time on patent issues over the years.

The GPL (regardless of its version) is not a protective shield against patents. It's not like someone attacks you with a patent, you say "but my program is GPL'd", and the aggressor has to retreat. Absolutely nothing like that. As a matter of fact, a lot of major corporations including Amazon.com, HTC, Salesforce.com, TomTom and who-knows-how-many-others have agreed to pay royalties for licenses to use patents that read on Linux.

So at the most one can argue -- but not in Google's case here -- that the GPL would help in the sense of an alleged infringer being able to point to a license granted to the relevant patents by their holder.

There's no way Google could impress Oracle by claiming it built Dalvik (had it done so) as a fork of IcedTea. IcedTea is a Red Hat fork of code Oracle/Sun made available under the GPLv2. But the implicit patent grant under the GPLv2 does not help if modified code infringes the same patent.

Google would only have been safe if it had used unmodified stuff published by Oracle/Sun, such as the OpenJDK or phoneME. In that case, Google could claim that by making such software available, Oracle/Sun granted a license to all of the relevant patents (so long as the code isn't altered). That, however, is not a special trait of the GPL (in whichever version). If Oracle/Sun had elected to publish stuff under the Apache license, Google's position wouldn't be weaker (arguably even stronger).

There is also a general patent promise related to the Java specifications. Bruce Perens highlighted that one. But that promise, again, has nothing to do with the GPL. It's a license-neutral patent grant that rewards (under whatever license) total compliance with the specs. It forbids supersets and subsets.

It calls into question the FSF's true motives and its defense of the principle of "software freedom" that it should effectively argue against the right to enhance or optimize such a standard as Java. Actually, it's very cynical of the FSF to claim the GPL protects "everyone's freedom" when the issue here is that the freedom to depart from the specs is denied by the aforementioned specs-related and license-neutral patent grant.

It's worth nothing that Dan Ravicher (then counsel to the FSF and now with the Software Freedom Law Center) pointed out back in 2004 that the GPLv2 doesn't really create legal certainty concerning the use of patents.

When evaluating the question of whether Oracle/Sun would have granted a patent license to Google in a hypothetical scenario, you always have to ask yourself: did Oracle/Sun publish something that Google used "as is" (without alterations)? With IcedTea, this test fails at the first hurdle: IcedTea itself was not published by Oracle/Sun itself. The FSF presumably mentioned IcedTea in order to try to reassure users of IcedTea that Oracle wouldn't go after them. The problem is that the moment IcedTea departs from the official Java specs, Oracle could use patents against it,cand the GPL wouldn't effectively prevent it from doing so if the code Oracle/Sun released under GPLv2 was modified. That fact was also recognized by the European Commission in its decision on the Oracle-Sun merger.

Where I do, however, agree with the FSF is that Google's preference for the Apache license is part of a strategy to enable Android-based products with essential proprietary/closed-source components, and that Google doesn't oppose the patentability of software. However, the FSF could say the same about many other companies, and doesn't. Some of those are known to fund the SFLC.

You can't have your cake and eat it, too

drag — Sat, 11 Sep 2010 06:07:42 +0000

> There was no way for Sun to release a flavor of Java under the GPL while imposing an additional condition as to the particular family of devices on which it may or may not run.

They could if they believed that the 'GPLv2 implicant patent license' is rather weak.

You can't have your cake and eat it, too

qubit — Fri, 10 Sep 2010 19:01:37 +0000

No. Not for phones. Sun/Oracle maintained that you had to use a particular flavor of Java for phones that was not the full JRE that you could run on servers or desktops.

Sun/Oracle can't have it both ways. There was no way for Sun to release a flavor of Java under the GPL while imposing an additional condition as to the particular family of devices on which it may or may not run.

Quoting from the GPlv2:

The act of running the Program is not restricted...

and

GNU GENERAL PUBLIC LICENSE...Version 2, June 1991...
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.

I find it very interesting that I've heard this statement repeated over and over again -- about how Sun "wouldn't allow" the so-called full version of Java to be run on phones. It's possible that Java originally had such a restriction before it was released under a Free Software license, but I believe the actual truth of the matter is that the full version of Java was just too much for small devices of the era to handle!

What Oracle Java stuff is GPLv2 anyway?

coriordan — Fri, 10 Sep 2010 16:54:10 +0000

I'm no expert, but I don't think that would be a problem. I think the "do what you want" wording of the GPL would foreclose any claims of a "this is for use on a desktop" limitation.

But yeh, it's hard to be sure with patent law...

But the good news is that I dug around and found that Oracle also distribute Java ME under GPLv2 (not sure if it's included in OpenJDK but it's definitely in Oracle's GPLv2'd "PhoneME"), so Google could use that and do "Java stuff on a mobile phone".

Actually, it would be very useful to make a list of exactly what Java specifications and packages and versions are distributed by Oracle under GPLv2:

http://en.swpat.org/wiki/Java_and_patents#What_exactly_wa...