It is my understanding that Sun's lawyers originally created the CDDL to be GPL-incompatible, with the specific desire that code under that license should not be able to aid their then-competitor, Linux. How Canonical can suddenly decide that Sun's lawyers were wrong about this, I don't know. I suspect wishful thinking.
I believe that Bryan Cantrill has stated that it was surprising to the people at Sun when outside parties didn't start incorporating ZFS and other technologies into Linux, because the original intention when the CDDL was written was not to be GPL-incompatible.
I'm going to go with the actual author of the licence (Danese), and not just because she authored it but because the idea that Sun management, seeing themselves struggle against Linux, would allow Linux to use their prized ZFS and DTrace, and thus offer even less reason to choose Solaris over Linux is simply insane from any kind of business point.
And not even now do we have anyone arguing here that CDDL is not GPL-incompatible, people accept that and Linus refuses to merge any CDDL code in to mainline, what is being discussed here is if ZFS counts as a derivative (and thus has to be GPL compatible) when it runs as a kernel module in Linux kernel space.
If that was the intent, is it not something that can be clarified and fix the problem? Or did they screw up too badly and it's unambiguously incompatible no matter what they say?
That is easy to disprove. If CDDL did not care, then one could follow the conditions in GPLv2 and the two licenses would be compatible.
License incompatibility can only occur if two license has conditions which are mutually exclusive. In this case, CDDL demands that source code must be distributed only under the terms of CDDL, and GPLv2 require that the distributed source code is made available under GPLv2. You can not comply with one without breaking the other, and thus we have license incompatibility.
> That is easy to disprove. If CDDL did not care, then one could follow the conditions in GPLv2 and the two licenses would be compatible.
The entire point of the CDDL is that it's module local. The only way to achieve GPL compatibility is to become the GPL. That would be impossible for the CDDL to do.
Yeah. If there's an unintended license-compatibility problem, the people who'd need to fix it are Oracle's lawyers. They might have a different attitude towards Open Source than Sun, to put it mildly.
Wouldn't a clarification of intent be based on evidence from the people that created the license? It doesn't matter what the lawyers think, they can't change history. You'd need the lawyers to relicense, a completely different way to solve the issue.
Why would Oracle spend any time correcting it? If you haven't figured out by now Oracle won't spend money on anything they don't think will make them money, I'm not sure what to tell you. And clarifying this doesn't seem like a path that would make them any amount of money.
I doubt that every single person that could clarify is employed by Oracle in 2016.
But my understanding after reading everything else is that the license is blatantly incompatible and there is no possible clarification that could fix it.
As a counterpoint, here's a 2011 video of Bryan Cantrill claiming that it is "emphatically not the case" that the CDDL was specifically developed to be GPL-incompatible: https://youtu.be/-zRN7XLCRhc?t=1375
It may not have been the "original intention" of the license, for whatever value of "original intention" that means, but it certainly was used for exactly this purpose - to keep the code out of Linux. Whether or not it was just a "happy accident" the CDDL was not GPL compatible, or whether it was engineered to be that way, I believe is mostly a moot point. (Though I'm still nowhere near believing Bryan Cantrill on this point. The executives and lawyers could have easily done this without the software developers involved ever knowing until after the fact.)
The GPL incompatibility was a well known fact before both of those tools were released, and it's frankly unbelievable this point wasn't raised internally at Sun before OpenSolaris walked out the door, given how much public discussion went into the license after it was first announced. The Sun executives knew exactly what they were doing when they chose that specific license for distributing D-Trace and ZFS.
I think all concerned (executives, lawyers, engineers) were simply trying to do the right thing by our source base, Sun's customers and the budding OpenSolaris community -- we really weren't trying to placate (or alienate) the Linux community, and didn't really factor in how the Linux community perceived its own license.
As for OpenSolaris, there were a couple of constraints with respect to the GPL: first, we didn't want to dual license (i.e., GPL + something else), because that was a mess that could lead to a license-based fork -- something we definitely wanted to avoid. Second, there was great concern about simply using the GPL because of exactly the ambiguity that's highlighted here: in 2005, we had many independent hardware vendors that had entirely proprietary Solaris drivers -- it was important to us that those be allowed to stay proprietary. We all liked the BSD license, but we also liked the idea of a weak copyleft -- and in that regard, the license that best captured what we were after was the MPL, despite some (minor) issues too that needed to be cleaned up. In the end, this is what we did (the CDDL is really just a cleaned up MPL) -- but (to be clear, and for what feels like the millionth time) we did it because of what we were trying to do for our own software and community, not because we were somehow trying to punish Linux.
As I've said before, one of my most vivid memories of the whole process was immediately after we launched OpenSolaris, standing in the conference room where we had just hit the carriage return, chatting with Chris Nadan from Sun Legal. We were both wondering aloud how long it would take for DTrace to show up in Linux -- but not from the perspective of license compatibility: we both accepted as a given that Linux would either view the licenses as compatible enough or wouldn't care about any incompatibility (after all, this is years after NVIDIA's binary drivers). Our only discussion was how much members of the Linux community would care about DTrace and how difficult the port would be technically. I remember saying that "maybe it will be two months, and maybe as long as two years." Of course, it was only a month or so later that SystemTap emerged from Red Hat -- and it became clear that Not-Invented-Here Syndrome would prove to be a much stronger force than any technical desire to assimilate DTrace or ZFS into Linux...
The NIH syndrome by established Linux developers definitely slowed down Linux's adoption of these technologies. However, the inability of Sun's marketing to promulgate how awesome its technologies were to the next generation of developers slowed down adoption even further. In my case, information about how awesome ZFS is did not reach me until about 5 years ago (I started contributing to the Linux port soon afterward) while information about how awesome DTrace is did not reach me until about 2 or 3 years ago.
I was barely even aware that a company called Sun existed until about 8.5 years ago when I was given access to a shell account on Solaris 9 at my university for a class on C at age 18. At the time, there was nothing to tell me that ZFS and DTrace even existed, much less how awesome they are. Back then, I had spent my entire life stuck in the desolate wasteland that is Windows computing with a tiny bit of LAMP sprinkled into my life that I had used to play webmaster of a Pokemon fan site between ages 10 and 15. Years later, I realized that I really disliked writing PHP.
Maybe things would have been different if Sun had tried fighting Microsoft for the PC market, but the knowledge that anything better out there existed just did not make it to me. I did not learn about Sun's technologies until I started actively looking for good technologies in their respective categories. I found out not long after my searches began, but it would have been nice had there been a bit more advertising.
These days, I cannot imagine going back to computing without ZFS and while I do not quite have DTrace available to me yet, when I finally do, I imagine that I will feel the same about DTrace too.
> It is my understanding that Sun's lawyers originally created the CDDL to be GPL-incompatible, with the specific desire that code under that license should not be able to aid their then-competitor, Linux.
Schwartz (or Sun under Schwartz) was not a fan of the GPL. As far as I remember they were hoping that Linux and OpenSolaris would together somehow migrate to an imaginary GPLv3 that never happened.
QUOTE: "Schwartz singled out the GPL provision that says source code may be mixed with other code only if the other code also is governed by the GPL. That provision is intended to create a body of software that must remain liberated from proprietary constraints. But Schwartz said that some people he's spoken to dislike it because it precludes them from using open-source software as a foundation for proprietary projects."
