LWN: Comments on "LLVM contemplates relicensing"

OpenBSD

Wol — Fri, 23 Sep 2016 01:06:24 +0000

Well, when I asked a solicitor to make a *minor* update to my will, and he deleted a *major* beneficiary in the process ...

Copying pre-existing text accurately surely isn't that hard?

Cheers,
Wol

OpenBSD

pabs — Thu, 22 Sep 2016 13:10:58 +0000

How does he propose to handle patent nonsense?

OpenBSD

epa — Thu, 22 Sep 2016 09:43:28 +0000

I understood that Theo wasn't making a legal statement that patent clauses were not 'valid' in copyright licences, but rather was expressing his personal preference that he wants a traditional set of 'copying conditions' for the program, dealing with copyright only, and any patent nonsense can be handled separately. Which I kind of sympathize with.

OpenBSD

louie — Thu, 22 Sep 2016 03:37:26 +0000

For the record, Theo's argument about patent clauses not being valid in copyright licenses is the kind of lawyering-by-engineers that drives lawyers nuts. It is superficially correct, but guess what, copyright law says very little about direct payment either. Or trademark. Or any of about a billion other things that routinely are handled alongside copyright in copyright deals.

"If you don't care about your software's use in OpenBSD..."

Which is increasingly the dominant position, in part because of stuff like this.

OpenBSD

epa — Wed, 21 Sep 2016 15:08:10 +0000

The lawyer's professional opinion doesn't determine what gets into OpenBSD. Theo's does. I would give similar weight to the views of Debian's licensing committee or Slackware's Grand Pooh-Bah of Licence Acceptability or whoever. Of course an expert in the field is more likely to be correct (though as I said, it very much depends on the exact question that was asked) but in the end it's not about who is right, but about finding some acceptable answer that most people can work with. If you don't care about your software's use in OpenBSD, you can disregard Theo's view on the matter, whether or not you think he is likely to be 'right' or 'wrong'.

OpenBSD

gowen — Wed, 21 Sep 2016 12:44:53 +0000

I've had to correct the work of pretty much every lawyer I've had the (mis)fortune to deal with.

You're brilliant and everyone else is incompetent. Got it.

OpenBSD

Wol — Wed, 21 Sep 2016 12:12:36 +0000

Give that choice, my money would be on Theo!

IME, far too many people are incompetent at their alleged field of expertise. I've had to correct the work of pretty much every lawyer I've had the (mis)fortune to deal with.

And the main cause of my occasional spats with PJ was over her blind faith in lawyers ...

Cheers,
Wol

OpenBSD

gowen — Wed, 21 Sep 2016 11:04:52 +0000

Oh, lawyers definitely disagree with one another (and are rarely, if ever, as certain as Theo always is). But if you are asking me to choose between an experienced IP lawyer's professional opinion and a software engineer's 6-sentence hand-wave legal argument, I know where the smart money goes.

OpenBSD

rahulsundaram — Wed, 21 Sep 2016 06:01:38 +0000

> Someone can always fork and add stuff under a different license.

Sure but that is not an incentive for the project itself to dual license. The rationale for doing so is very weak here.

OpenBSD

Otus — Wed, 21 Sep 2016 05:22:58 +0000

> > But that applies with any permissive licence
>
> Yes. Hence dual licensing between two different versions of the same license is problematic.

The point is that licensing under any permissive-enough license is already as problematic. Someone can always fork and add stuff under a different license.

OpenBSD

rahulsundaram — Tue, 20 Sep 2016 19:30:02 +0000

> But that applies with any permissive licence

Yes. Hence dual licensing between two different versions of the same license is problematic.

OpenBSD

epa — Tue, 20 Sep 2016 18:34:26 +0000

the possibility of a fork under one of the licenses, resulting in a legal block that prevents code from being shared by the forks.

But that applies with any permissive licence (if the licence forbids this kind of fork, it can't be said to be permissive).

LLVM contemplates relicensing

epa — Tue, 20 Sep 2016 18:32:41 +0000

Apache has a built-in CLA.

Don't you require a signed statement from the contributor that they are licensing their work under the Apache licence? Or is the licence document in a distributed tarball considered sufficient? (Since they might claim that they had intended to use a different licence for their modified version, as they are within their rights to do.) What I'm getting at is whether there is any real reduction in paperwork.

LLVM contemplates relicensing

dberlin — Tue, 20 Sep 2016 14:39:34 +0000

It was considered.
It would have required separate legal agreements to handle contributions to ensure the proper rights grants were in place (IE CLA's of some sort) to be able to do that.

Apache has a built-in CLA.

Plus, BSD still would have needed an exception anyway, as it requires reproducing the notice in binary distributions :)

OpenBSD

rahulsundaram — Tue, 20 Sep 2016 12:12:41 +0000

>In this case 'the trouble of dual licensing' doesn't seem to be an issue, since all the code is already available under the old licence, so there is no auditing or permission needed to keep offering it as an alternative.

That's just one factor in dual licensing troubles. There are many other problems including the possibility of a fork under one of the licenses, resulting in a legal block that prevents code from being shared by the forks. I think you are estimating the complications of dual licensing.

LLVM contemplates relicensing

epa — Tue, 20 Sep 2016 12:09:02 +0000

It would be possible to use the BSD licence for copyright, with a separate licence granting rights to use software patents. In some ways that would be cleaner and simpler.

OpenBSD

epa — Tue, 20 Sep 2016 11:41:03 +0000

If the newer license is better, it makes sense to only offer under the newer license.

Indeed, but the point I'm trying to make here is that 'better' can be a matter of opinion, rather than fact (or "my lawyer says so"), and given that opinions tend to differ, it makes sense to allow for the possibility that the old licence might be better for some, even if in your personal view there is no reason to prefer it.

If the licence change has some other motivation like stronger copyleft, then of course it makes sense to drop the old one (the FSF didn't continue to dual-license under GPLv2 when v3 came out). But if ostensibly you want to have permissive terms, the wiser choice is not to assume that everyone shares your (or your lawyer's) view that the new licence is just as free and just as permissive as the old.

In this case 'the trouble of dual licensing' doesn't seem to be an issue, since all the code is already available under the old licence, so there is no auditing or permission needed to keep offering it as an alternative.

OpenBSD

rahulsundaram — Tue, 20 Sep 2016 11:13:54 +0000

> If the claim is that the new licence really is just as good as the old, and grants everyone just as many rights (plus a bit more), then it should be effectively a no-op to offer the software under a choice of the old and new licences

I don't see why anyone would do that. If the newer license is better, it makes sense to only offer under the newer license. Remember that many licenses including MPL or Creative Commons licenses have automatic upgrade clauses. Just because some developers don't want to go through the trouble of dual licensing for very marginal benefits does not validate OpenBSD's perspective here although if they want to be dogmatic about only accepting BSD and not accepting other permissive licenses including Apache, that's certainly their choice.

OpenBSD

epa — Tue, 20 Sep 2016 10:55:32 +0000

Nonetheless, lawyers do disagree with each other, and the answer you get very much depends on how you pose the question. And if OpenBSD's stance is partly "patent terms do not belong in a licence, it should cover copyright only", then we can discuss whether this is a wise policy to hold, but it's not something that depends on the view of any particular legal advisor.

If the claim is that the new licence really is just as good as the old, and grants everyone just as many rights (plus a bit more), then it should be effectively a no-op to offer the software under a choice of the old and new licences. If there is some reason why the old licence is discontinued (for future releases) at the same time as the new licence is introduced, that seems to suggest the OpenBSD people may have a point.

OpenBSD

gowen — Tue, 20 Sep 2016 08:28:15 +0000

Classic Engineers Disease from Theo there. Play a juvenile semantic game and act like that's the final word on a legal issue. Unsurprisingly, this reflex legal opinion does not seem to be held by the actual lawyer that LLVM Foundation consulted.

OpenBSD

epa — Tue, 20 Sep 2016 06:27:05 +0000

I know that OpenBSD is allergic to the Apache licence, and forked Apache for this reason. (Mailing list thread at http://marc.info/?l=openbsd-misc&m=107726777116056&... but I can't find the top of the thread - I miss Gmane...)

There is a bit of a disconnect between those who insist that the new Apache licence is equivalent to BSD/X11 with some 'extra protections', and those who think that this extra stuff makes it no longer free or at least no longer a BSD-style permissive licence. It might be better to dual-license the software under both Apache and a plain permissive licence, so those who insist on simplicity and feel they can do without the extra patent 'protection' can continue to distribute the software.

LLVM contemplates relicensing

Otus — Tue, 20 Sep 2016 05:22:25 +0000

How about LGPLv2.1? I thought Apache 2.0 was incompatible with it too? That is not a problem?

LLVM contemplates relicensing

pabs — Tue, 20 Sep 2016 05:04:29 +0000

That is a bit of an icky exception...

LLVM contemplates relicensing

xtifr — Mon, 19 Sep 2016 23:27:50 +0000

Ah, they want one license for both the compiler and the runtime library. Ok, that makes sense. Thanks.

Still a little curious about the warranty thing, but my main question is definitely answered.

LLVM contemplates relicensing

edgewood — Mon, 19 Sep 2016 21:22:07 +0000

The BSD license doesn't cover patents that go along with the code. Patent protection for users and contributors is in the list of goals for the relicensing. Given those goals, BSD isn't an appropriate license.

LLVM contemplates relicensing

ballombe — Mon, 19 Sep 2016 20:56:11 +0000

> Seems eminently sensible.

But is it ? The BSD license is a simple permissive license while the Apache 2.0 (not the 1.0) is complex and ten time as long and not even GPL2 compatible.

Welcome in Europe

sytoka — Mon, 19 Sep 2016 19:29:15 +0000

Some compagnies like to paid for nothing ! You alway have dream vendor and dream client ;-)

PS : USA people have to help Europeen citizen to keep our region free of patent... Move all your fondation here could help us ... and you !

Welcome in Europe

pizza — Mon, 19 Sep 2016 18:37:10 +0000

There are quite a few patents granted on software in many European countries, despite it nominally not being allowed.

Welcome in Europe

sytoka — Mon, 19 Sep 2016 18:28:38 +0000

No stupid patent in Europe on code... Move your fondation here, fight agains't patent in USA !

You are welcome in a libre country ;-)

LLVM contemplates relicensing

excors — Mon, 19 Sep 2016 18:09:30 +0000

And currently the runtime libraries are dual-licensed as MIT, so applications linking to the libraries today don't need to provide a copy of any license. But having different licences for the compiler and runtime is problematic when LLVM developers want to move code from compiler to runtime, so it seems sensible to have a single licence for everything, and the proposed change will achieve that without placing new requirements on applications.

LLVM contemplates relicensing

josh — Mon, 19 Sep 2016 17:57:39 +0000

Without that exception, anyone who built their software using clang, or who linked in one of the LLVM project's runtime libraries, would need to include a copy of the license. That's a little too broad for a compiler runtime library or for incidental code included by compiler intrinsics. That would even affect binaries built by a project whose upstream didn't plan for building with clang/LLVM.

LLVM contemplates relicensing

xtifr — Mon, 19 Sep 2016 17:25:46 +0000

Seems eminently sensible. IANAL, but the described solution is more-or-less the one that popped into my head when I read the problem description.

The only thing that really puzzles me is why they want to add an exception for APL clause 4.a, "[y]ou must give any other recipients of the Work or Derivative Works a copy of this License." The old license required including all three parts of the old license with binary distributions. (See the second bullet point via the link in the article.) Not bundling the license with binaries seems to be a new thing they want to introduce going forward. Are there really people who object to that?

At minimum, I'd be a little concerned that the standard disclaimers found in all these licenses no longer have to be included. But, as I said before, IANAL, so I don't really know whether that's a problem or not.