SmartOS (based on IllumOS) released

SmartOS (based on IllumOS) released - with KVM

Posted Aug 16, 2011 9:54 UTC (Tue) by nye (guest, #51576)
In reply to: SmartOS (based on IllumOS) released - with KVM by rodgerd
Parent article: SmartOS (based on IllumOS) released - with KVM

>"I've decided I can use GPL code because I've written my license to say it's OK and doesn't taint my codebase" would be called out as bullshit in no time flat if it was Microsoft; I see no reason why some random OpenSolaris derivative should get a free pass

Because it's far from clear that a kernel should be legally considered a derivative work of some module that you build for it. In fact, their position sounds pretty safe.

Thought experiment: if somebody writes a kernel driver for Windows and releases it under the GPL does that mean we should 'call [it] out as bullshit'? Same thing.

to post comments

SmartOS (based on IllumOS) released - with KVM

Posted Aug 16, 2011 12:11 UTC (Tue) by pboddie (guest, #50784) [Link]

Because it's far from clear that a kernel should be legally considered a derivative work of some module that you build for it.

You have to work back from the objectives of the GPL to understand what the licence is trying to do. The aim of the GPL is to have a software system whose complete code exhibits the property of offering the four freedoms. When the licence talks about a derived work, it attempts to use established terminology to entertain the possibility that a piece of software integrated with another comprises a system that, like the GPL-licensed portion, should provide those four freedoms.

(Note to the usual people complaining about the FSF "forcing" people to pick their licence: the consequence is that the "other" portion of such a system need only be available under a compatible licence, and the author of that work need only care about the matter if they are doing the combining, anyway. Thus, the Solaris people can just say, "No, you're not supposed to do that - stop it!")

That's why someone pointed out that choosing one's own definition or understanding of "derived work" doesn't necessarily get anyone off the hook. It's very possible that someone interpreting the licence could arrive at an interpretation that suggests a more far-reaching definition, unless some explicit exception has been stated by the GPL-licensed work's authors.

SmartOS (based on IllumOS) released - with KVM

Posted Aug 16, 2011 12:51 UTC (Tue) by paulj (subscriber, #341) [Link] (6 responses)

The Solaris/IllumOS kernel does not become derived of the Linux KVM driver, of course. However the resulting combined work of IllumOS + Linux KVM surely is derived of BOTH those works, and so any distribution in that form surely must be licensed by each of the copyright holders.

SmartOS (based on IllumOS) released - with KVM

Posted Aug 16, 2011 13:54 UTC (Tue) by nye (guest, #51576) [Link] (5 responses)

>However the resulting combined work of IllumOS + Linux KVM surely is derived of BOTH those works

Your 'surely' here is not universally agreed upon - both IllumOS and KVM were developed independently of each other and neither requires the other. On that basis it's hard to see why either would be considered a derivative of the other just because they happen to be used together.

I expect your argument is that modifying KVM to make it work with IllumOS *makes* it a combined work derivative of both (that would be my argument in your place, anyway). That seems reasonable, but it *also* seems reasonable to conclude that bundling the two together with some modification necessary for technical reasons of interoperability is insufficient for these two separate works to now be considered one single work derivative of both.

I'm not saying 'there is definitely no problem', but I am saying that the situation isn't as clear as you make out, and a sound argument could be made that there is no single combined derivative work.

SmartOS (based on IllumOS) released - with KVM

Posted Aug 16, 2011 14:17 UTC (Tue) by paulj (subscriber, #341) [Link] (4 responses)

They are, separately, independent works, of course. However the KVM announced here is not the stock Linux KVM, but one ported (with significant effort) to work within Illumos. I thought that was reasonably obvious from the blog post. ;) I.e. it is a derived work of Linux KVM and Illumos.

To quote the blog post: "just getting it to compile in any meaningful form would take significant effort", "KVM is so tightly integrated into the Linux kernel", "hard to know when to pull in the Linux implementation of a particular facility or function versus writing our own or making a call to the illumos equivalent (if any)". The blog post seems to be fairly explicit about the fact that much judgement, work and creativity was needed to solve a number of problems. It's very hard to think software copyright wouldn't apply to the apparently substantial amount of code they say they had to modify, re-write and write.

My understanding is that the bar to a work being derived is pretty low. I.e. small patches with obvious, no-other-way fixes might not be worthy of copyright, but if there's any room for creativity then that implies that the work does fall under copyright. Further, when you have a one body of code dependent on a very *specific* other piece of code, then my understanding is that the dependent code derives from the other. The Illumos KVM patch seems, by the blog post, to depend on both the Linux KVM and IllumOS code.

That said, the CDDL and the GPL are both free software licenses, and if all the copyright holders concerned are all happy, then there needn't be a problem.

SmartOS (based on IllumOS) released - with KVM

Posted Aug 16, 2011 18:40 UTC (Tue) by fuhchee (guest, #40059) [Link] (3 responses)

"I thought that was reasonably obvious from the blog post. ;) I.e. it is a derived work of Linux KVM and Illumos."

Why "... and Illumos" ?

SmartOS (based on IllumOS) released - with KVM

Posted Aug 16, 2011 20:47 UTC (Tue) by jone (guest, #62596) [Link] (2 responses)

Perhaps that should really read:

"it is a derived work of Linux, KVM, and Illumos"

realistically KVM (or the KVM modules more specifically) is a derived (and dependent) work from interfaces in the Linux kernel, and the Joyent port is a derived work of certain Linux interfaces (some they chose not to port), KVM interfaces, and the Illumos kernel .. (of course there are derived works in the linux kernel from earlier sun code covered now by the CDDL - but who's counting?)

there's a good talk that bmc did (is doing?) for linux-kvm 2011 forum this week that goes into much more detail on what was involved .. slides here:
http://www.linux-kvm.org/wiki/images/7/71/2011-forum-port...

SmartOS (based on IllumOS) released - with KVM

Posted Aug 17, 2011 15:21 UTC (Wed) by vonbrand (guest, #4458) [Link]

The story told there is that this KVM is a straight derivative from the original code, and thus GPLv2 only (look at the code's notices in Linux). Very bad...

SmartOS (based on IllumOS) released - with KVM

Posted Aug 18, 2011 18:31 UTC (Thu) by hummassa (guest, #307) [Link]

There is no "derived work from interfaces". "Abstraction, Filtration, Comparison" -- Google it, please. If there was such thing, Linux would be a derived work from Unix, remember the SCO case? "errno.h" etc...

SmartOS (based on IllumOS) released - with KVM

Posted Aug 27, 2011 3:50 UTC (Sat) by cas (guest, #52554) [Link] (1 responses)

Thought experiment: if somebody writes a kernel driver for Windows and releases it under the GPL does that mean we should 'call [it] out as bullshit'? Same thing.

If Microsoft distributed such a driver with windows, then they would be bringing their entire kernel under the terms of the GPL. For some strange reason, this is unlikely to ever happen.

If a hardware manufacturer did that, they'd be breaking the terms of the driver's GPL....unless they were the author, or got the author to give them permission to do that - e.g. "this driver code is GPL but you're also allowed to distribute it on new PCs with Windows, as long as you abide by the terms of the GPL as it applies to this driver". Similar to the GPL + KDE special permission we used to see back when KDE had a GPL-incompatible license.

if the driver is distributed by itself for the end-user to install on their own machine, then there's no license problem at all. The GPL does not restrict usage, only redistribution.

SmartOS (based on IllumOS) released - with KVM

Posted Aug 29, 2011 13:27 UTC (Mon) by anselm (subscriber, #2796) [Link]

If Microsoft distributed such a driver with windows, then they would be bringing their entire kernel under the terms of the GPL.

No, they wouldn't. They would be distributing that driver without a valid license, which is a different thing altogether.

Wishful thinking notwithstanding, a piece of software will never automatically be GPLed in its entirety, simply because the manufacturer incorporated some GPL code into it and distributed the result. (Incidentally, this is why the GPL is not a »virus« GPL »infection« doesn't work that way.) The only problem in this case is that the GPL code is not distributed according to its license. This means that (a) the manufacturer may be liable to the author(s) of the GPL code for damages resulting from past unlawful distribution of that code, and (b) the manufacturer needs to do something to avoid further non-compliance with the GPL in the future.

They can of course solve the problem by voluntarily putting their codebase under the GPL, too, but if we're talking »big proprietary package« it is more likely that they will simply remove the offending GPL code and replace it by something with a more tractable license, or make arrangements with the copyright holder to be able to use the code under a different, non-GPL, license.