Re: and calls kernel API's directly is a derivative work

Posted Feb 19, 2016 14:20 UTC (Fri) by pizza (subscriber, #46)
In reply to: Re: and calls kernel API's directly is a derivative work by drag
Parent article: Kirkland: ZFS licensing and Linux

> Also it's worth pointing out there that lack of mainstream support of ZFS on Linux is a example of how destructive 'copyleft' licenses can be and can negatively impact technical decision making. GPL and CDDL are both strong 'Free Software' licenses and ZFS and Linux are both Free software projects, yet they are both isolated from one another by their respective copyright restrictions. That is not to say that 'copyleft' is bad, but only that it does have it's price.

This isn't a problem with copyleft; it's a problem due to two different licenses having incompatible terms.

And it's worth mentioning again that the CDDL was *deliberately* designed to be incompatible with the GPL -- for the specific purpose of keeping ZFS out of Linux. [1] This, coupled with Oracle's demonstrated aggressiveness over copyright matters, really makes me wonder how Canonical can reasonably conclude that acting against both the letter and spirit of the licenses is a remotely sane path to pursue.

[1] https://en.wikipedia.org/wiki/Common_Development_and_Dist...

Re: and calls kernel API's directly is a derivative work

Posted Feb 20, 2016 2:22 UTC (Sat) by rahvin (guest, #16953) [Link] (2 responses)

Don't blame Oracle for the CDDL, that was Sun's doing. They wanted to make sure Solaris had something Linux didn't, so they wrote a deliberately incompatible license. The simple fact is Sun did this before Oracle even thought about purchasing Sun and it was entirely deliberate. This is the reason up until this that no one dared ship ZFS as part of a Linux Distribution and all the code, modules and even packages were outside the main distribution channels.

Personally I think if Oracle decides they don't want this they will kick Canonical's butt in court.

Re: and calls kernel API's directly is a derivative work

Posted Feb 20, 2016 14:57 UTC (Sat) by pizza (subscriber, #46) [Link]

> Don't blame Oracle for the CDDL, that was Sun's doing.

Oh, that wasn't my intent, and I completely agree with what you wrote.

But it is worth pointing out that Oracle could chose to re-license ZFS if they were so inclined. Personally that's the outcome I'm hoping for (ZFS is frickin' awesome!).

Re: Don't blame Oracle for the CDDL, that was Sun's doing

Posted Feb 22, 2016 21:40 UTC (Mon) by ldo (guest, #40946) [Link]

Yes, we can blame Oracle for the CDDL. Because they have assimilated Sun, and all its copyrights. They could rescind the CDDL if they wanted to, and replace it with something sane, but they choose not to.

Re: and calls kernel API's directly is a derivative work

Posted Feb 21, 2016 12:31 UTC (Sun) by paulj (subscriber, #341) [Link] (4 responses)

And it's worth mentioning again that the CDDL was *deliberately* designed to be incompatible with the GPL

This has been gone over a good few times here before on LWN, with the likes of Bryan Cantrill (influential Solaris engineer) replying even. I won't rehash, but it has to be said that the CDDL most definitely could not have been chosen just to be incompatible with the GPL. Not even Danese Cooper said that, her claim was that it was "partially because".

I wasn't privy to the licence discussions, but my mentor at Sun was. And I'm pretty confident Simon Phipps' description of things is much more representative. There were many factors involved in choosing a licence, and all the ones I know about had nothing to do with Linux. Even if it were true that some involved wanted to make it harder for code to get into Linux, that doesn't change the fact there were several other factors that would still have prevented Sun choosing the GPLv2. In particular, the lack of defensive patent protections clauses, the desire to not impose open-sourcing requirements on 3rd party ISVs (Sun didn't want to require driver writers to have to open their code), etc.

Further, if using the licence to deny ZFS and DTrace and what not to Linux was the goal, then the CDDL of itself does not accomplish that. The CDDL itself is pretty much fine with being incorporated with other works as long as the CDDLed portions stay CDDLed. Authors of GPLed works can just give themselves an exception to include CDDLed works - the licence goals of both are broadly similar too (copyleft licences).

"Ah, but there's far too many people involved in Linux, so it could never realistically grant itself an exception; don't even know who they all are!". Well, Linus *did* unilaterally change the licence at least once. Further, other big projects have relicensed even while not being able to contact every copyright holder (can't remember the project right now).

Anyway, what I know for sure is that, at worst, Linux compatibility was but 1 of a number of concerns that led to Sun choosing the CDDL, and that the wholly-non-Linux-related concerns would also have ruled out choosing the GPLv2.

Re: and calls kernel API's directly is a derivative work

Posted Feb 21, 2016 17:09 UTC (Sun) by raven667 (subscriber, #5198) [Link] (2 responses)

> just to be incompatible with the GPL. Not even Danese Cooper said that, her claim was that it was "partially because".
> [...]
> Anyway, what I know for sure is that, at worst, Linux compatibility was but 1 of a number of concerns that led to Sun choosing the CDDL

Wait, what is the purpose of all these words, you are conceeding that you agree that GPL incompatabily was a requirement in the crafting of the CDDL, but the point you want to add is that this wasn't literally the _only_ license criteria, a point that was not under serious criticism? It seems like the real difference in opinion is not in the facts but in how much you personally value them, the deliberate incompatabily is a bigger deal for some people than for you.

Re: and calls kernel API's directly is a derivative work

Posted Feb 21, 2016 20:07 UTC (Sun) by paulj (subscriber, #341) [Link] (1 responses)

I didn't concede that at all. I never heard of Linux incompatibility being a motivating factor, and I had a chat with a mentor of mine about this who was involved (as well as a brief chat with the director of Solaris engineering in a 1:1, when he visited our site).

Danese is the only person who has said that, and she said "partially".

All I wanted to do was point out that there were other reasons why Sun would not have chosen the GPLv2. E.g., one being that a lot of Solaris engineers (inc. the former engineers in management) were much more BSD-aligned and just anti-GPL - nothing to do with Linux, just reflecting a long-standing, often dogmatic, split in the free Unix world between BSD and FSF aligned camps; the other being that Sun legal wanted stronger patent lawsuit defences than the GPLv2 provides. All I'm saying is that Simon Phipps' account rings much more true to me.

AFAIK, regardless of the truth behind Danese's comment, the outcome was just not to be one where Solaris would have been GPLv2 (or compatible licensed) - independently of the Linux question.

Basically, the conspiracy theories on this are way overblown.

If it was said it was cause there was a visceral dislike for the GPL amongst many inside Solaris eng and Sun legal and management were never going to agree to a BSD licence, then - that'd ring a lot more true.

Re: and calls kernel API's directly is a derivative work

Posted Feb 21, 2016 21:28 UTC (Sun) by paulj (subscriber, #341) [Link]

Oh, and I've gone and listened to that Debian workshop video on licenses to hear Daneses exact words, and she did *not* mention Linux, she said the "Mozilla [licence] was selected partially because it is GPL incompatible".

So the comments here saying it was because of /Linux/ incompatibility are twisting Danese's words somewhat.

Where she said that there were people from engineering in the discussion pushing very hard for the BSD licence, and some for the GPL - that's exactly what I heard from my mentor (I suspect my mentor would have been one of the ones who had pushed for GPL). And Simon Phipps at ~36 minutes says the same thing - that the Solaris engineering community wanted the BSD licence. Phipps goes on to say the "we" (I guess himself, Danese, ??) didn't want that, because they wanted a copyleft licence.

Next, you have to know Sun was already using a Mozilla based licence; that Sun (e.g. Phipps) didn't want to use a strong copyleft licence (didn't want to force ISVs to have to copyleft); that Sun wanted strong patent protections for itself and other community members. Phipps also states (38m) that engineering would have "quit" if they'd picked GPLv2.

That last bit is consistent with my limited experiences. And it had nothing to do with Linux. It had to do with very strong BSD v GPL licensing views within Solaris engineering, and Solaris engineering having very strong Unix/BSD roots.

Remember, the engineers - the ones who largely ruled out the GPLv2 as a valid choice - would have chosen BSD otherwise! So it just could _not_ have been about screwing Linux, cause if they'd got their way, the code would have been under a licence that was perfectly compatible (one way) with going into Linux!

As per previous, I think the GPL was ruled out for Solaris more because of bias around the licence, and ancient BSD/FSF divides, than anything to do with Linux directly.

Re: and calls kernel API's directly is a derivative work

Posted Feb 25, 2016 1:33 UTC (Thu) by vonbrand (subscriber, #4458) [Link]

Oh, come on. When Linus changed the license from the original "no commercial use" to GPLv2 he was still the person who had written almost all of the kernel himself, with help of perhaps a half dozen close collaborators.