Debian and Nexenta collide [LWN.net]

Debian and Nexenta collide

Posted Nov 10, 2005 0:47 UTC (Thu) by paulj (subscriber, #341) [Link]

Actually, he did more than call it "ethical", he said the CDDL "is a free software licence" (with the reservations i mentioned regarding the GPL incompatibilities). RMS is not one to call something "free" lightly.

Regarding patents, yes, they're confined to the CDDL code implementing the patent. Quite obviously the CDDL can not give liberal patent grants if it wishes to be able to specify patent MAD and a patent commons amongst stakeholders (users/developers) in a CDDL project. Note that the patent grants implied by the CDDL *are* available to GPL code, indeed code under any licence. It is the *GPL* which restricts itself, simply because it is quite old now and doesn't deal with the question in any way except to say essentially the patents must be licenced unconditionally to everyone, royalty-free. Note that this is obviously *not* how the GPL licences software copyright (you must follow the restrictions of the GPL), and it's quite probably *not* how the GPLv3 will approach software patents. In other words, watch out: don't today condemn the CDDL for trying to use patents as a lever in the interests of Free Software, as the GPLv2 already does with copyright, and then tomorrow applaud the GPLv3 when it may well try very similar tricks[1].

A really simple way forward would be for the GPLv3 to recognise the validity of the CDDL patent-pooling/MAD clauses and allow GPLv3 applications to avail of any patent grants in CDDL code, by allowing GPLv3 code to link to such CDDL code. There is *nothing* in the CDDL to stop the FSF, or anyone else, doing so (as long you follow the CDDL wrt to the CDDL /files/, and any derivatives thereof).

I work for Sun, so I guess I'm tainted and part of the Sun conspiracy, so don't believe me - believe RMS.

1. This is pure speculation on my part, but it has been reported that tackling software patents is on the agenda for the GPLv3 and I personally think the CDDL patent clauses would be quite suitable for the next revision of the GPL.

Debian and Nexenta collide

Posted Nov 10, 2005 1:11 UTC (Thu) by cventers (subscriber, #31465) [Link]

Did you read the article Stallman wrote? The one called "Sun's no-op
announcement" that referred specifically to the patent grants in
question? Unless I'm drunk, or he's totally wrong, then Sun's CDDL patent
grants extend to nothing except Sun's own CDDL-released software.

>> So what has really happened here? Reading the announcement clearly, I
>> think that it doesn't announce anything at all. It simply describes,
>> in a different and grandiose way, the previously announced release of
>> the Solaris source code as free software under Sun's idiosyncratic
>> license, the CDDL. Outside Solaris, few or no free software packages
>> use that license--and Sun has not said it won't sue us for
>> implementing the same techniques in our own free software.

Patent grants implied by the CDDL are *certainly* not available to GPL
code. Read OpenSolaris.org's comments:

>> The CDDL provides an explicit patent license for code released under
>> the license. This means that you can use, modify, and redistribute
>> code released under CDDL without worrying about any patents that the
>> contributors of the code (including Sun) might have on the contributed
>> technology. The license also includes a provision to discourage patent
>> litigation against developers, by revoking the rights to the code for
>> anyone initiating a patent claim against a developer regarding code
>> they have contributed.

Found on ZDNet:

>> Perhaps the most interesting among the list of commentators is IBM's
>> vice president of standards, Bob Sutor, who in his blog entry,
>> commended Sun for embracing the spirit of open source, but then added
>> "As far as I understand it, Sun is not pledging these patents for use
>> in any open source project as [IBM] did."

So how can you say:

>> Note that the patent grants implied by the CDDL *are* available to GPL
>> code, indeed code under any licence. It is the *GPL* which restricts
>> itself, simply because it is quite old now and doesn't deal with the
>> question in any way except to say essentially the patents must be
>> licenced unconditionally to everyone, royalty-free.

Please clarify! Better yet, please tell your bosses at Sun to clarify in
public!

Debian and Nexenta collide

Posted Nov 10, 2005 2:29 UTC (Thu) by paulj (subscriber, #341) [Link]

I did, and yes the grants extend only to the CDDL-released software concerned. However, my point was that the CDDL allows any software to make use of it - that includes GPL software. The only thing the GPL software author has to do is grant herself an exception to link her own code to the CDDL code implementing the patent, and hey presto, she's got a patent licence (by using the CDDL code).

RMS had a concern that Suns' grant was not royalty-free, non-exclusive (AIUI) and hence does not extend to GPL software. The thing though is that Sun can not have the CDDL setup a Free Software patent pool if it also grants such liberal rights (you could avoid the pooling conditions of the CDDL by simply availing of the RFNE grant then). That the GPLv2 is old, needs updating and doesn't provide for patent-pooling is hardly a drawback of the CDDL ;).

Finally, there's a simple solution, the GPLv3 can just avail itself of that pool, by allowing linkage or inclusion of CDDL code (ie, all the GPLv3 really has to do is to just recognise the CDDL as compatible for purposes of GPL - no more).

As both an FSF supporter and a Sun bod, I very much hope the GPLv3 will do that.

Debian and Nexenta collide

Posted Nov 10, 2005 3:18 UTC (Thu) by cventers (subscriber, #31465) [Link]

As of yet I've resisted letting your affiliation with Sun have any impact
on my replies. What I will say is that your attitude...

>> That the GPLv2 is old, needs updating and doesn't provide for
>> patent-pooling is hardly a drawback of the CDDL ;).

...is typical of Sun Microsystems and is the reason why I suspect you
won't ever be real open-source competition for Linux. Sun isn't the god
that it thinks it is.

Your assertion that GPL software could link CDDL software and enjoy its
patent is specifically identified in the CDDL as false:

>> (d) Notwithstanding Section 2.1(b) above, no patent license is
>> granted: (1) for code that You delete from the Original Software, or
>> (2) for infringements caused by: (i) the modification of the Original
>> Software, or (ii) the combination of the Original Software with other
>> software or devices.

I'll take the GPL over the CDDL any day. Thankfully, I suspect that the
silent majority agrees with me.

Debian and Nexenta collide

Posted Nov 10, 2005 4:41 UTC (Thu) by paulj (subscriber, #341) [Link]

>> That the GPLv2 is old, needs updating and doesn't provide for
>> patent-pooling is hardly a drawback of the CDDL ;).

...is typical of Sun Microsystems and is the reason why I suspect you won't ever be real open-source competition for Linux. Sun isn't the god that it thinks it is.

Odd, the above is a view held by the FSF - nowt to do with Sun. One of the greatest challenges today is what to do about the climate of patent-trolls, particularly in software. The CDDL attempts to do something about it, the GPLv3 almost certainly will attempt to introduce wording to tackle patent-risks too. As an FSF supporter, I certainly hope so anyway. The above was not intended a comment as a Sun employee, more with my Joe Bloggs (FSF supporter) hat on if anything. Sorry if came across as snide.

Your assertion that GPL software could link CDDL software and enjoy its patent is specifically identified in the CDDL as false:

"(d) Notwithstanding Section 2.1(b) above, no patent license is granted: (1) for code that You delete from the Original Software, or (2) for infringements caused by: (i) the modification of the Original Software, or (ii) the combination of the Original Software with other software or devices."

I don't actually know if my assertion is false or not. I *think* is true, but my assertion carries no weight - hope that's obvious. However, with respect to this 2.1d clause, note the "Notwithstanding", hence clearly 2.1d does not trump 2.1b. 2.1d therefore, by my layman's interpretation, is there to state the obvious:

"You can't infringe on Suns' patents in your own code and claim you have a patent licence through 2b simply by linking to, or wedging your infringing code in, with Suns' CDDL code".

In the latter case, you'd have to CDDL your code anyway, or breach the copyright. Note that even if you infringe on a patent of Sun, that still does not mean Sun are going to sue you, it is entirely at the patent holders discretion whether to sue or not. There are lots of things that can be done to resolve infringement other than suing people, e.g. you may be contributing to OpenSolaris, it's quite unlikely Sun would then sue you. Sun does not have a record of running around suing people, further if it's quite inconsequential Sun can easily just ignore it, at its discretion. Patent law is not like trademark law, your claims are not affected by ignoring infringement (AIUI).

(Any large corporation which does R&D almost certainly already has lots of patents in its portfolio which are being infringed here and there. By and large, corporations don't care as long as they're not sued themselves - only then does the portfolio come out ;). That is the current climate, for better or worse.)

I'll take the GPL over the CDDL any day. Thankfully, I suspect that the silent majority agrees with me.

You can use what you want. Sun uses the GPL too. Most of my work to date at Sun has been on GPLed software. Ditto for quite a few other Sun engineers. As for the CDDL, its fundamentals are fine - RMS has stated this too.

Whether it would have been possible to find the wording to allow the patent grants to follow modified works, I don't know. From a reply of yours in another thread, you obviously think that wording should have been put there. I'd agree in principle, but have doubts about whether it could be done without resulting in a vague and unclear licence.

Anyway, it's late, plus I still have to read the weekly edition! Night.

Notwithstanding

Posted Nov 10, 2005 6:36 UTC (Thu) by ncm (subscriber, #165) [Link]

"Notwithstanding" doesn't mean what you seem to think it means. Better look it up.

Notwithstanding

Posted Nov 10, 2005 8:22 UTC (Thu) by paulj (subscriber, #341) [Link]

Strangely enough, I did look it up to make sure it meant what I've always thought it meant and (AFAICT) it does. Care to expand?

Notwithstanding

Posted Nov 10, 2005 16:44 UTC (Thu) by vmole (guest, #111) [Link]

It means that 2.1(d) completely overrides (or perhaps "limits") subsection 2.1(b). That means that the patent grants do NOT apply to any modified code, or code that incorporates CDDL code in new ways. That may not be the intent, but that's what it says.

Notwithstanding

Posted Nov 10, 2005 23:03 UTC (Thu) by paulj (subscriber, #341) [Link]

Interesting, that goes against completely against the sense of "Notwithstanding" that I've understood. The "notwithstanding" refers to some obstacle that may limit or make more difficult the effect/action being referred to. I had thought till now at least.

Notwithstanding

Posted Nov 10, 2005 16:50 UTC (Thu) by rfunk (subscriber, #4054) [Link]

A simpler version: No matter (and despite) what 2.1b says, 2.1d applies.

This is NOT about interfaces!

Posted Nov 10, 2005 1:56 UTC (Thu) by paulj (subscriber, #341) [Link]

And this all is relevant... exactly how ?

Well because that clause may or may not be enforceable. If there's copyright "boundary" (there likely is) then it might be harder for the GPL to reach across it. But that'd be something you'd have to argue before a judge and jury.

Let's forget about interfaces, GPL and CDDL for a bit. Let's talk about two writers: Justina and Max. Both created some great stories and now publisher want to publish single book with stories by Justina and Max. Justina granted right to publish stories in any form to everyone but Max does only grant this to anyone who will not publish Justina's creations in the same book as his storis. Can our publisher then publish this book ? Of course not: you need permissions from all authors! Note: stories are totally unrelated - the only connection is single bookcover (and empty piece of paper is hardly copyrightable).

The problem is that that's likely not the some thing. In the case you describe, the publisher almost certainly has contracts with these author. If an author has a clause in their contract that makes their copyright licence conditional on not publishing with the work of some other author, then it is *CONTRACT* law which gives that clause weight, not copyright - contract law is far more flexible.

A licence under copyright alone is more difficult to have mandate restrictions not relating to copyright - you can *grant* rights with a unilateral licence, but you possibly can not take away rights over which you have no claim through copyright, for otherwise copyright licences could make all sorts of silly claims. You can simply disregard such clauses, for they have no effect on copyright (you can't break a copyright licence if copyright is not involved). Least, it used to be that way, until stuff like UCITA came along to start giving shrink-wrap copyright licences more contract-like powers, at least in the USA.

As a thought experiment, imagine if the GPL had a clause stating that you could not distribute GPL works if you also used MS Windows, or if you had a dog. I.e. think of some silly clause which would be legal under contract law (not breaching discrimination laws, etc.) yet has nothing to do with act of copying the work concerned. It's not too hard to think such clauses would be unenforceable, given the right judge ;). In the case of this clause, even more so when the licence *does* allow the act of distributing a GPL work without the OS components it depends on, the contentious components here.

It poses problems both way really. If its not enforceable, that's a problem for any non-copyright related /restrictive/ clauses in licences like the CDDL and GPL. There aren't many such clauses though, but this clause is one. If it's enforceable, then obviously yes there's a theoretical legal and moral problem for Nexenta GNU/OpenSolaris.

However, that's all quite speculative and there's also a decent chance my reasoning is flawed. ;) So lets just assume the clause is quite valid from now on. Where do things stand then? And isn't this a storm in a tea-cup?

Nexenta have a licence to distribute OpenSolaris (the CDDL). They have, other than this complication, a licence to distribute GPL packages. Further, the GPL *does* allow GPL software to depend on non-GPL OS components, even where they are *completely* proprietary, (OpenSolaris is "Free", if not GPL). It also allows GPL software to be distributed in aggregate with non-GPL works, even completely closed proprietary works.

So we're having a big argument because one Free application is being distributed with a Free OS, because the licence of the Free application has an unfortunate oversight in it which neither the licences author nor the author of the application probably ever intended to constrain such distribution.

"May be GPLv3 will solve this problem, may be not - but right now Nexenta is in quite real trouble."

How exactly? They're not in trouble if either:

a) (ideally) The dpkg copyright holder(s) grant a CDDL exception for dpkg. Or even just clearly state "hey, its perfectly ok by us" as an official position - even that would be sufficient legally under the British legal systems in US, UK and elsewhere.

b) (far less good, if only from a moral perspective) The dpkg copyright holder(s) don't sue.

And *no one* but the dpkg copyright holders can sue btw. Their opinion is all that matters. (I not sure what, if any, status those posts on those debian lists have wrt expressing the opinion of the dpkg copyright holders)

If a sensible solution can not found then that means a Free OS and Free applications essentially can not be distributed together, which I think most sides would agree is not a desirable outcome.

This is NOT about interfaces!

Posted Nov 10, 2005 4:34 UTC (Thu) by kimoto (subscriber, #5244) [Link]

And *no one* but the dpkg copyright holders can sue btw. Their opinion is all that matters. (I not sure what, if any, status those posts on those debian lists have wrt expressing the opinion of the dpkg copyright holders)

At least one dpkg author is hostile to linking dpkg with a CDDL libc.

$ zcat /usr/share/doc/dpkg/changelog{.Debian,}.gz | grep -F doogie@debian.org | wc -l

345

This is NOT about interfaces!

Posted Nov 10, 2005 4:34 UTC (Thu) by AJWM (guest, #15888) [Link]

You were so close ... and then you missed it.

But, the license is the only thing that gives you permission under copyright law to distribute copies. A clause stating that "you may only distribute if you do not own a dog" is no less enforceable than a clause saying "you may only distribute if you pay the author one thousand dollars". Giving somebody money has "nothing to do with act of copying the work concerned", directly, but it is certainly the basis of most licenses to copy.

A copyright holder may be as capricious as he chooses (bar certain antidiscrimination laws) about what he wants in return for granting permission to copy his work -- you have no right otherwise to copy it.

Of course any judge might make an arbitrary, erroneous decision -- which is why there are appeals courts. Unless the clause on which a license grant was conditioned were actually illegal, there's no way that a decision which effectively vacates the copyright holder's rights and wishes would be upheld.

(ob: IANMALTAEPH - I am no more a lawyer than anyone else posting here)

This is NOT about interfaces!

Posted Nov 10, 2005 8:34 UTC (Thu) by khim (subscriber, #9252) [Link]

Hey, he said: given the right judge. At this point it does not matter what GPL says, what CDDL says - given the right judge you can get any resolution writtable in English.

This is NOT about interfaces!

Posted Nov 10, 2005 9:09 UTC (Thu) by paulj (subscriber, #341) [Link]

A copyright holder may be as capricious as he chooses

Hmm, well with further research it does indeed appear that I'm completely wrong in thinking that a licence can't mandate arbitrary conditions (as long as they are legal). Cheers ;).

This is NOT about interfaces!

Posted Nov 11, 2005 6:31 UTC (Fri) by hppnq (subscriber, #14462) [Link]

One of the nice things about the GPL is that it is actually quite easy to grasp. The license itself explains why it exists, and how it can be applied.

Which, unfortunately, does not mean that people are any less confused by it. Some parts are of course a bit hazy, but I suppose it's got more to do with the fact that people tend to cherish their own set of definitions and notions, even if they don't realize it.

Consider the extremely simple concepts of copyright and copyleft and how both are abused by both sides in discussions like these: lots of knee-jerking. What's all the more surprising, is that these are definitions, it's not like we've discovered a new law of nature that we can't fully understand yet.

PS: paulj, of course I'm not trying to ridicule you by suggesting how simple all this is, I'd much rather salute you for admitting you were, errr, not completely right -- and flexible enough to adjust your point of view.

This is NOT about interfaces!

Posted Nov 18, 2005 3:22 UTC (Fri) by robilad (guest, #27163) [Link]

I'd hope that the morality of a deed trumps the question wheather someone can be successfully sued. One might get away with doing something, but that does not make it a nice thing to do.

cheers,
dalibor topic

This is NOT about interfaces!

Posted Nov 10, 2005 9:20 UTC (Thu) by Duncan (guest, #6647) [Link]

> [A]ll distributions who actually ship
> NVIDIA's binary driver on the same CD
> as linux kernel are in trouble: they
> are violating GPL - for exactly the
> same reason. Nobody tried to sue them
> yet (AFAIK, anyway) but it's clear
> copyright violation.

They haven't been sued, but they are under notice from some of the kernel
hackers to cease and desist or face legal action.

See the LWN kernel page quote of the week for the Oct. 26 edition (quoting
Greg K-H). There's quite a discussion there as well, including a couple
rather long "expositions" on the legal details of the GPL and how they
relate to binary kernel modules, by yours truly. =8^)

http://lwn.net/Articles/157225/

Duncan

Sounds like co-distribution is perfectly legal

Posted Nov 10, 2005 15:39 UTC (Thu) by hazelsct (guest, #3659) [Link]

Now back to Nexenta, CDDL and GPL. As we've seen from sample above it does not matter of libraries are using the same interface - if they are published as single product (and Nexenta presumably wants to do exactly this) then licenses must be compatible. In fact this was reason for GPLv2 ! Note this part of it: In addition, mere aggregation of another work not based on the Program with the Program (or with a work based on the Program) on a volume of a storage or distribution medium does not bring the other work under the scope of this License. Why such exception was ever needed ? Easy: without this exception SUSE (with GPLed bash and proprietary Macromedia Flash) will be illegal! GPLv1 had no such provisions and this was a problem GPLv2 solved. May be GPLv3 will solve this problem, may be not - but right now Nexenta is in quite real trouble.

But aren't you disproving your own point here? Since the GPL expressly allows distributing non-GPL software in the same product, it would seem to allow distribution of GPL software with CDDL libc.

Furthermore, Sun and other proprietary unices have been distributing GPL gcc with proprietary libc implementations for literally decades without infringing, or at least without infringement suits. Why the sudden storm in a teacup over distribution with free software?

(Incidentally, this is not the same as the old KDE/Qt flap, which was about distributing GPL apps linked against non-free non-system libraries with non-standard interfaces, and nearly sank the original Corel Linux distribution -- or at least its apt/KDE software installation tool.)

As a Debian developer myself, I have to agree with paulj that the complaint of these DDs is a counterproductive restriction on free software, let alone childish infighting between free software camps which will more than likely damage the reputation of the Debian project and the community as a whole. To one day say, "SCO is wrong, interfaces are not copyrightable", and the next day turn around and say, "You'd better not link my free software to your free system software with a standard interface!" is truly pathetic.

Sounds like co-distribution is perfectly legal

Posted Nov 10, 2005 16:53 UTC (Thu) by vmole (guest, #111) [Link]

You've missed the distinction between "mere aggregation" and "derived work". The claim is that when you distribute (say) dpkg with the Solaris libc, you are distributing a derived work combining the dpkg code and the libc code, because (this particular binary) of dpkg will not work without (this particular binary) libc. Remember API!=ABI, and by including the header files provided by the Solaris libc, you're including Solaris libc specific code.

Sounds like co-distribution is perfectly legal

Posted Nov 10, 2005 17:12 UTC (Thu) by khim (subscriber, #9252) [Link]

But aren't you disproving your own point here? Since the GPL expressly allows distributing non-GPL software in the same product, it would seem to allow distribution of GPL software with CDDL libc.

Not really. GPL allows "mere aggregation" of two programs on the same medium regardless of their license if they are not related (relevant part os quoted in previous post).

This is other question. Why FSF never sued Apple and/or Sun ? May be they had no money or they were not sure it's a good idea. Who knows. Copyright owner is not forced to sue anyone who violates copyright (that's not trademark).

I agree that all this problem is quite silly and looks like storm in a teacup but this does not make it less real.

Sounds like co-distribution is perfectly legal

Posted Nov 10, 2005 20:54 UTC (Thu) by vonbrand (subscriber, #4458) [Link]

Oh, come on now. Before there even was Linux (and thus the possibility to even build a fully GPL-compatible system), all GPLed software had to run on propietary systems, simply because there was nothing else to be had. Before Linux, glibc was a sad joke, and everything used the propietary libraries that came with the OS. Standard joke was that the first task with a new Sun was GNU > /usr/local.

So now there is Linux. And glibc works. And what I did before to our Suns will not be tolerated anymore, just because?! I just can't see the difference between shipping gcc compiled for Solaris, and doing the same with dpkg. AFAICS (IANAL!) this kind of stuff is explicitly allowed by GPLv2.

Sounds like co-distribution is perfectly legal

Posted Nov 11, 2005 4:46 UTC (Fri) by kimoto (subscriber, #5244) [Link]

One rationale from Thomas Bushnell for why the funny wording is in the GPL2 (which was designed in the age of proprietary systems):

The special exception allows you to ship, for example, emacs binaries linked against the proprietary HPUX libraries, provided HP distributes those libraries along with the major components of HPUX (that is, they cannot have unbundled them), and provided you are not shipping those libraries yourself. This is specifically designed to prevent HP from including an emacs binary which is linked against their libraries, shipping the whole thing as part of HPUX, and not providing the source for their libraries.

Sounds like co-distribution is perfectly legal

Posted Nov 11, 2005 11:09 UTC (Fri) by khim (subscriber, #9252) [Link]

Oh, come on now. Before there even was Linux (and thus the possibility to even build a fully GPL-compatible system), all GPLed software had to run on propietary systems, simply because there was nothing else to be had.

Yes, but was GPLed software included with proprietary systems ? I know of one sample from that era: NEXTStep (later Mac OS X). All others shipped without any GPL components. If end-user does install GPL software on OS - this is one story (there are special exception in GPL), but if vendor ship GPLed component as part of base system... this is different story.

Sounds like co-distribution is perfectly legal

Posted Nov 11, 2005 13:12 UTC (Fri) by vonbrand (subscriber, #4458) [Link]

There sure were. DG-UX (Data General) used gcc as their stock compiler. The propietary (I believe source wasn't available) Pro toolset from Cygnus was built around gcc, gdb, binutils, etc. Not just some "shipping with" there, without those tools the rest made no sense at all.

Besides, this is more than a bit strange, all around: DEC provided a tape with "extraofficial software" with their systems, Sun later shipped a CD with such stuff. I'm sure other Unix vendors did likewise. There are vendor sites giving GPLed stuff compiled and packed for their systems (including whatever parts of the libraries and such this requires). And now, when the same thing is being done with an OS shipped in source form, under a somewhat free license (much freer that original Solaris or Ultrix, in any case),it is bad?

Sounds like co-distribution is perfectly legal

Posted Nov 13, 2005 20:28 UTC (Sun) by Ross (subscriber, #4065) [Link]

The same thing with NextStep. GCC was the system compiler. They modified it to support ObjectiveC, and were forced to release the source code to the ObjectiveC front end, but not for the whole operating system.