Debian and Nexenta collide

Posted Nov 10, 2005 8:14 UTC (Thu) by paulj (subscriber, #341) [Link]

Thank you for being fair about addressing my points. I really don't want to write as if LWN is my platform to flame you or Sun but I feel as if my concerns are legitimate, and I suspect many feel the same.

And same from me. This is LWN after all (and a locked article too!), we should be quite capable of lively, but reasonable, discussion. :). Thanks btw for opening my eyes a bit to other viable interpretations of some of the clauses we had discussed. I hadn't considered them (and on one point at least so far I stand corrected, thanks (patent grants extending to modified works)).

I think we can agree that the wording in the CDDL is questionable and possibly open to interpretation (more so than either of us would like).

I wouldn't say questionable. I agree it is less than ideal. However, as you can well imagine, the licencing issue had a *lot* of discussion internally at a high-level involving engineering, legal and executives. A lot of very smart people from those branches were involved in debating and thinking through the issues. I wasn't one of them (I'm a newbie junior staffer ;) ), but am quite confident there must have been good reasons for the wording as it is. My educated guess is that there wasn't a good way to word the patent grant broadly without making it very vague and diluted. Better to keep it tight and clear - if better wording is found in future, the licence can be modified. The flip-side is that you might not be able to undo over-broad patent grants if you make a mistake through vague wording. An engineer's approach to software-licencing perhaps.

The remaining difference then, it seems, is that you trust Sun whereas I'm not completely inclined to.

I do, but I can't convince you with that, other than asking you to take that in faith. I would ask though that you apply Occam's Razor sometimes, e.g. in deciding whether Sun have bad intentions in not granting broad patent rights or whether instead Sun simply wanted to keep the licence simple rather than risk unintended consequences with ill-defined wording to grant more liberal patent rights, consider whether maybe the latter explanation is the more simple and hence maybe the more likely.

Why couldn't Sun have chosen to explicitly, and in plain english, license all of its patents to any CDDL-licensed product derived from the original? Why would it care to keep these patents locked up?

Well, I really don't know to be honest. I can try to guess, from my POV as someone inclined to be quite sympathetic to Sun (and aware of the internal corporate culture). I would say firstly that there is the point above, it simply is easier to keep a licence 'tight' initially and to add more grants in later revisions if needs be, than the other way around. Also Sun considers itself responsible enough to be able to provide some stewardship over these patents and copyrights, goes without saying. Further, Sun is very much trying to devolve governance of OpenSolaris to a community consensus model. There is a OpenSolaris CAB in place, three of its five members are external to Sun. Formulating the governance process for OpenSolaris is still a work in progress, part of that process is figuring out how to specify the relationship between Sun and OpenSolaris in terms of stewardship, custodial duties, rights and responsibilities. Read the draft, it should give some flavour at least of the direction OpenSolaris might take. It's a learning process obviously though.

It's not anything to do with MAD, because Sun demonstrated (to my glee) with OpenDocument that there are other approaches, by licensing their relevant patents to all OpenDocument implementations except when an adversary attempts to attack OpenDocument with their own patents.

Hmm, but that is exactly patent MAD though. Also, ODT is a document format, not an entire Unix kernel and userland implementation. Not inconceivable there'd be different considerations.

my suggestion would have been to simply use the GPL. The GPL has some big advantages. Bob Young of Red Hat pointed one out in particular in a recent interview: the GPL is well understood.

ACK. Like I said, the GPL was a strong contender (I think one of the people involved in the opening of Solaris mentioned that in a blog shortly after the OpenSolaris launch). But, as mentioned in another post, there were reasons not to. E.g. the GPL does not allow linkage to/from closed code, an issue for drivers. Then there's the fact that the GPL does not (yet) try mitigate patent risks.

Sun may have not been an abuser in the past, but plenty of people have been burned by the phrase 'trust me'.

The main issue you've brought up though is patents, how exactly would the GPL have helped you have more trust wrt patent claims? Remember, the copyright holder doesn't have to follow the GPL, so the GPL does not stop them giving you their code under the GPL and then later asserting patent claims against you. You actually have a bit more protection from Sun suing You (in the CDDL "You" sense) with the CDDL than you would have if Sun had GPLed OpenSolaris. I find it a bit strange that this would be a problem ;). So Sun should have gone further by giving You *no* patent grants at all?

Again, I'm not trying to be "smart" or snide - I'm trying to understand where you're coming from, by getting you to expand on those parts of your position I find difficult to understand at present.

Sun responded:
>> "Instead of sniping from the outlands of participation in the open
>> source community it'd be real nice to see HP try and take a run at
>> dislodging Sun from the number one slot as contributor of code among
>> commercial companies.

So HP take PR pot-shots at Sun, and we respond. What's wrong with that exactly? They're a competitor of ours, they do their best to compete with us, we do our best to compete with them. FWIW, the point made by Sun above is not without merit. Sun hackers contribute to lots of stuff, both as part of their official Sun duties, and in their own time.

Sun's behavior has been adversarial and confusing.

Adversial, sure. Sun are a company, Sun has a duty to compete :). RedHats' executives have also taken PR pot-shots at us, and we at them. I guess that's part of what executives are supposed to do, I don't know. ;)

Confusing, I'll accept that actually (not speaking for Sun). I think Suns' "message" (to use some marketing speak) has been a bit mixed over the years. If you were kind, you might chalk that down to a company in the process of grappling with how to best adapt to new business realities of an open-source world. I think Sun has, in the last two years or so, finally decided on exactly what it needed to do, and the "message" should have been clearer since then (and stay clear).

Sun refuses to fully open Java

I'm not in that part of Sun, but again, you can bet there were a lot of discussions about it. In fairness, the specifications for the various parts of a JVM are open, and you can implement a compatible JVM if you wish. I don't really know enough about the issues here to give any useful comment though, sorry.

half-opened Solaris

Hmm, well, there's a lot to Solaris. I think Sun did mention it would take time and would be progressive, and that there'd be some bits which Sun simply could not release due to 3rd party interests, which would eventually have to be rewritten from scratch. See the CAB link above too.

So if you don't have inherent trust for Sun, what obvious reasons might Sun have for creating the CDDL? Either they didn't want anyone to be able to mix Solaris code into Linux, or they wanted extra control over what happens to Solaris in the "open" world, or perhaps even both. It doesn't seem genuine, it doesn't read to me like real open source, and I don't trust it.

Well, I hope then you're not using Mozilla in any shape or form, an even more untrustworthy licence than the CDDL! Seriously though, Sun didn't create the CDDL, they cleaned up the MPL. Mixing Solaris code into Linux? Well, they're quite different kernels, code simply is not going to be a straight-swap, nowhere close. Concepts from the Solaris kernel, those are not covered by copyright anyway. E.g. Linux has its own implementation of the Solaris slab allocator, quite a while already anyway (described in a USENIX paper by Jeff Bonwick a long time ago).

If instead you mean GNU/Linux, the userland applications, well yes of course you can swap (core utilities and libraries excepted, they tend to be kernel specific anyway). Why couldn't you? Linux and Solaris have *long* been running the same sets of applications anyway. E.g. You'd be very hard-pressed initially to tell whether you're logging into JDS on a Linux core, or on a Solaris core OS.

That's part of the beauty of Unix, that we have this independence of application from OS (not always perfectly, but better than any other system). Its open and good and allows for both strong cross-pollination and competition, in technical ideas and even in other areas like development models. Linux no doubt has learned things from Solaris, as it has from BSD and other systems. Solaris no doubt has things to learn from Linux, as Sun does from the community models built up around Linux. Indeed, one of the meetings CAB held was to investigate the models and processes these other systems used and to get hackers from those other communities to explain a bit about to them to the OpenSolaris CAB.

As a Linux user, you don't have to trust or like Sun at all, but maybe you don't have to be quite so suspicious of them. ;)

Regards,

Paul Jakma