Luis VillaRamblings on software, law, and the spaces in between.

The FSF got some things wrong in this process as well. (But I had dinner in between posts.)

• The FSF has failed to convince people that patents are a threat. The FSF folks apparently believe (1) that patents are a threat to free software projects and (2) that patents are immoral. These are distinct issues. I think if they’d focused on (1) they’d be ahead. Clearly the kernel guys think the best way to cope with the patent threat is to ignore it. The FSF has not effectively countered that argument, and so the kernel folks (and lots of other people) think that v3′s patent clauses are unnecessary and political, instead of a necessary defense against a dire threat.
• The FSF has failed to convince people that DRM is a taking of freedom. For the kernel folks (and a lot of other people) DRM is just a choice programmers have made on top of their stack. I tend to disagree (for reasons I’ll probably make clear in perhaps another post this weekend) but however you slice it, the FSF has clearly failed in convincing people that DRM built on top of Linux is the equivalent of shackles built on top of their code, and hence, that the DRM clauses in v3 are necessary.
• I’m not sure if this is a failure or not, but it was pretty clear from the beginning that the consultation was going to be an open way to get feedback, but that Richard was going to make the final calls. Given how famously difficult he is to convince of anything, and how the big problems to be solved by the license were apparently chosen before any open consultation, it is pretty understandable that some people (like the l-k guys) were turned off, and have chosen to act outside the system instead of within it.

These are not minor problems, but I should note that these should not completely blot out all the good the FSF has done within the process. The committee process has been a very open one, where just about anyone can contribute in a meaningful way. (Hell, they let me on one of them.) The software used to track and collate comments has overall worked well, and made it easy to spot problem areas quickly. (Sadly, it was not used on a mission statement.) So overall, despite the big-picture PR mistakes made above, the process has been very good and very smooth so far. CC, and any other community pondering license changes, should be following along to learn how to conduct a license review process.

I just posted about what the kernel guys really want. They are saying some other non-substantive things, but most of these are either solvable or just plain wrong.

• Virtually everything they are saying about license proliferation is wrong, except for the obvious statement that license proliferation is bad. We’re all agreed on that. GPL v3 increases compatibility with other licenses (apache license would now be compatible, for example, helping solve all the hoops beagle has to jump through currently to use lucene), so gpl v3 helps this problem, not hurts it. They are working hard to make even more licenses compatible, if they can- lots of people are looking at making eclipse license and gplv3 compat, for example.
• Similarly, allowing GPL to add and remove clauses in an orderly way would actually reduce fragmentation, not increase it as they claim. Sure, you’d get lots of GPL variants- but many other licenses would quickly die away as they became GPL variants (LGPL, for example, will do just this), so the net number of licenses actively being used in new could should stay the same. And if there are 100 licenses, and 90% of them share 95% of their language, we’re way better off than the current situation- where we have 100 licenses, which mostly agree on shared goals, but which share no language and hence are hard to make work with each other. [The shared language things matter- if I find code under a GPL variant, it is a lot easier to convince someone to add or subtract one clause than it is to convince them to change from, say, MPL to GPL (any version.)]
• They criticize the wording of the patent provision- but some very, very large patent patent portfolio owners have had large teams of lawyers participate in this process, and they’ve been willing to work with the draft. So it seems likely that the patent clause is not nearly as hosed as they think it is. The real thing to watch here is IBM- they are both the biggest contributors to the kernel, and the largest holder of patents in the world. So if they remain involved, then there is a pretty good chance that they think that the patent clause is (at least) salvageable.
• They completely ignore the many small and real problems that have been solved- the license is now more likely to be upheld by international courts and has a clearer definition of linking, for example.
• Their complaints about the DRM language are… well, maybe FUD is a little strong, but saying ‘people were confused at the first draft, so we should give up altogether’ is pretty irritating. They should either say flat out ‘we disagree philosophically with the DRM clause’ (which is fine, and honest) or they should work with those who are working hard to clean up the language by giving specific complaints.
• c’mon, guys- when you say ‘politics’ we all know you mean ‘how we define ‘freedom”, or some variant thereof. Quit playing the linguistic game of calling it ‘politics’.
• [Later edit: I totally forgot, of course, the most serious disingenousness- the belief that the FSF wasn't political, and is now political. C'mon. The FSF has been political from day one, guys- everything they are doing now would have been pretty damn predictable from the first day RMS heard about DRM, and the first day someone asked 'what about this patent?' on an FSF mailing list. To claim that this is now betraying the trust of people is really pretty silly. Is it in disagreement with what most kernel FLOSS contributors believe? Quite possibly. Is it in disagreement with what the FSF has been advocating for more than a decade? No, not at all. C'mon.]

So yeah… again, a shorter, sweeter, more straightforward declaration discussion would have been preferable. They’ve got a strong core point- people disagree about what the core freedoms are. They don’t need to surround it with groundless claims to substantiate it.

The kernel guys just put out a doc about their feelings on GPL v3. It is really a remarkable document, if for no other reason than that 30-ish kernel developers actually agreed on something :) My thoughts on it are that they are saying some very substantive things, but that they are dodging the real discussion- and maybe it is time for us to have that discussion again.

What they are saying, substantively*:

• GPL v2 works just fine.
• No problems are being solved by the new license.
• The DRM clause limits what users can do, which contradicts the freedoms which users had under v2 and which might scare away both users and contributors.
• The patent clauses would also scare away potential contributors.
• The FSF is violating the trust of those who have used the GPL under the assumption that future versions would not be fundamentally different.

Freedom, and how it is defined, is really the issue that all these rotate around. How does it tie these together?

• GPL v2 works just fine, and no problems are being solved by the new license- it does work fine, if you think that people using your code to build DRM-enabled products is acceptable. Otherwise it is very good- but clearly flawed.
• The DRM and patent clauses limits users and scares away contributors- both of which are true, but assumes that DRM does not violate meaningful freedoms, and that patents are not a threat to our codebases. (Both of which are defensible positions, I believe, though I don’t agree.)
• FSF is violating our trust by being ‘political’- which is true, if you think that FSF has never been political. If you’re sane, and realize FSF has from day 1 been political, then you might disagree with this :)

What the kernel guys really seem to want is a GPL 2.1 which reflects their very pragmatic approach to open source- take, as long as you give back. Past that, do whatever. What the FSF- and particularly Richard- wants is a GPL which reflects the FSF’s long-term approach to free software, where patents and DRM are another restriction (like binary software) that could reduce freedoms. Nothing in GPL v3 is inconsistent with FSF’s very long-term mission- to increase what they see as Freedom.

So really, as far as I can see, this is the old Open Source v. Free Software debate coming to a head again, in yet another guise. (Worse, it might be the GPL v. BSD license debate, where everyone argues about the definition of ‘freedom’, all over again.) I really don’t know where I stand on this, but I think it is clear most of the kernel developers are clearly in the Open Source camp. That’s fine- it is certainly their right, and frankly, I think they represent a majority of GPL v2 users. I do wish they had been more forthcoming in this inevitably important document, and spent less time confusing the issue. If they had, the document could have been a lot shorter, a lot sweeter, and a lot more accurate.

Let me make a stab at it, in fact, honestly trying to be as neutral as possible on the definition of Freedom:

“We believe that our particular community of GPL v2 users have come over time to a different definition of freedom than the FSF. We believe that the focus of the GPL should be on the elements that encourage collaborative participation, which include simplicity, enforced code-sharing, and end-user freedom. We reject the FSF’s attempt to define freedom in such a way as to include charged issues like DRM and patents, which our contributors disagree greatly about. As a result, we call on the FSF to stop the current discussion about GPL v3 and create a GPL v2.1, which does not seek to expand the Freedoms of GPL v2 in controversial directions, and instead focuses on strengthening and clarifying the terms of the current GPL so that we can more securely defend the rights and freedoms we believe we currently have, and which we have created a large community around.”

So, yeah. That’s really what it took them five pages to say. I think, put that way, they’ve got a fairly sound argument- clearly lots of people would want to use a GPL 2.1 that had the same focus as GPL v2, but with more refined legal safeguards. I personally would prefer a GPL v3, that prevents people from using my code to create devices whose primary purpose is to deliver DRMd content, and which protects me from the patent portfolios of companies which I may like but can’t trust. But clearly that is a difference of opinion that merely having an open comment period is not going to paper over.

It is telling, of course, that the kernel guys who wrote the paper (and Linus) did not participate in the open comment period. They could easily have done so, but I think at some level they know that this really isn’t about an issue they can discuss with the FSF. They must know that there is no way in hell GPL v2.1 is coming out of the FSF- convincing Richard to redefine what he sees as freedom (and the threats to it) is not ever going to fly. So they didn’t participate- they did an end-run. I wish they hadn’t run so far out, though.

Linus (as usual) says it very clearly and straightforwardly, but again, to the echo chamber of LKML and not to the FSF directly:

“My personal opinion is that a lot of the public discussion has been drivenby people who are motivated by the politics of the discussion. So you have a lot of very vocal GPLv3 supporters. But I think that the people who actually end up doing a lot of the development are usually not as vocal, and haev actually not been heard very much at all.

In some sense, the poll is a way for the people who actually do a lot of the work to show that the FSF doesn’t speak for necessarily even a very big portion of actual developers.”

Where he says ‘politics’ read ‘freedom’ and… yup. Sounds pretty familiar.

* They are saying a lot of other things that frankly I think they got very wrong. Another post about that shortly.