Snapd 2.31 Better Supports Wayland Via Mir, Canonical Hires Another Mir Developer

[starshipeleven]

As you (and Luke) point out, everything released before the change in license would still be available for distribution after the change. So, as you also point out, even if Mir becomes very common, and then Canonical changes the license, the community could continue to rely on (and improve a fork of) whatever version of Mir that had been relied on before the change. Sure, the community would need to maintain a fork moving forward. But that's no different from any time that a maintainer walks away from a project, whether due to lack of interest, financial reasons, or otherwise.

It would allow them to pull an EEE, which is something that you can't do (as easily) with GPL, and is the entire point here. GPL is supposed to avoid EEE as any change has to be opensourced, while with permissive it's not the case.

Their CLA is basically a way to have their work be GPL for everyone but Canonical, that can treat it as permissive for their own use.

This way they can get a nice product but they would have stolen the work of others. And I say "stolen" because work published with GPL should not become closed source. If the original license was Apache or some other permissive then it would be totally fine as the writing was on the wall.

This puts them in a higher ground than others, that can't simply take the source and close it and make their own EEE on it.

Which is something Oracle did with their ZFS closed source fork, it has more (and interesting) features that make it superior to the open one. The ZFS license was permissive though, so what they did is not technically wrong, this is just an example of what could happen.

If Mir becomes very common, and then Canonical changes the license, but the community fails to create a fork of the last GPL version, well, . . . I don't foresee that happening. If it does happen, shame on the community.

This would require first that the community bites the bait. Which didn't (and won't) happen.

But what happens if a project wants to change its license for a socially beneficial reason, but is prevented from doing so? For example, although some dispute the claim, it has been claimed that Oracle would be willing to relicense ZFS under GPL, but is prevented from doing so because it doesn't have the legal right to do so.

Not relevant in this case. All Canonical projects were started from scratch and they could have used permissive licenses or pure GPL at any time if they wanted to.

But in any case, if, at some time in the future, Canonical feels that they can take Mir further under the ASL2.0 license (or a similar license, like BSD, MIT, etc.)

They won't. The entire point here is having more control over their projects than anyone else does, so they can pull EEEs or lockdown whenever it feels convenient to do so. Using a permissive license would defy that, as it would allow ANYONE to do the same.

Note that Google is using permissive licenses for Android (the OS, not the kernel which is GPL), not some wacko bullshit trickery with GPL and a CLA

[GizmoChicken]

In a previous post, I wrote:

If Mir becomes very common, and then Canonical changes the license, but the community fails to create a fork of the last GPL version, well, . . . I don't foresee that happening. If it does happen, shame on the community.

To which you responded:

This would require first that the community bites the bait. Which didn't (and won't) happen.

Great! Problem solved!! Now that Canonical's CLA is a non-issue, we can all move on and agree never to discuss Canonical CLA again. Yippee!!



No? Okay, let's continue…

In all seriousness, although I don’t have a crystal ball, I don’t foresee Canonical ever attempting to exercise their option to relicense Mir under a license other than GPL. If Canonical has plans to change the license for Mir, they likely would have done so from the start, which would have been easier and would avoid much legal uncertainty. No, my guess is that Canonical includes that provision in their CLA (which CLA includes several other notable provisions, including a provision that seeks to prevent a contributor from later suing the project for infringing that contributor's patent(s) covering the contribution) as way to keep their options open, “just in case” situations change in the future.

Nonetheless, I acknowledge that it is possible that Canonical may exercise their option to relicense Mir, and so, for the sake of argument, I’ll consider that possibility here.

You wrote:

You cannot change the license retroactively, so whatever was out before will remain out with the same license.

Agreed. And I wish that more people who post comments against Canonical's CLA realized this.

You also wrote:

The point here is that [Canonical's CLA] would allow Canonical to pull tricks like letting some components become very common everywhere (ala systemd) and then change license, causing massive havok and forcing people to do their bidding or face massive issues in either maintaining themselves the last opensource stack or migrate to something else.

Meanwhile, stuff that got very widespread, like systemd, has a plain GPL license, no CLA, no bullshit. Even playing field for all.

And you wrote:

[Canonical's CLA] would allow them to pull an EEE, which is something that you can't do (as easily) with GPL, and is the entire point here. GPL is supposed to avoid EEE as any change has to be opensourced, while with permissive it's not the case.

I think you would agree that X Window System has “become very common everywhere” and "widespread" on Linux desktops. As you know, the “X.Org Foundation has chosen [a modified version] of the MIT License as the preferred format for code included in the X Window System distribution.” As you also know, the MIT license provides legal permission to use MIT licensed code in closed source. Yet, no one has been able to “pull tricks” or “pull an EEE” with the X Window System in any way that has resulted in “causing massive havok and forcing people to do their bidding or face massive issues in either maintaining themselves the last opensource stack or migrate to something else.”

Assuming the GPL version of Mir ever becomes “very common everywhere” and "widespread" on Linux desktops (which even you doubt will happen), what leads you to think that Canonical could “pull an EEE” on the community? That is, even if Canonical were to later release a non-GPL version that had additional/improved features, why wouldn’t the community happily continue using the GPL version of Mir that had already “become very common everywhere” and "widespread" on Linux desktops?

If you fear that the community would acquiesce to Canonical merely because the community would be unwilling, or unable, to maintain the GPL version of Mir, that’s just silly. If a GPL version of Mir ever becomes “very common everywhere” and "widespread" on Linux desktops, because Canonical could go bankrupt or otherwise unable to support Mir, the community must be prepared to support a GPL fork of Mir. If the community isn’t able to support a fork of the GPL code that has “become very common everywhere,” shame on the community.

And you wrote:

[Canonical's] CLA is basically a way to have their work be GPL for everyone but Canonical, that can treat it as permissive for their own use.

This way they can get a nice product but they would have stolen the work of others. And I say "stolen" because work published with GPL should not become closed source. If the original license was Apache or some other permissive then it would be totally fine as the writing was on the wall. . . .

Note that Google is using permissive licenses for Android (the OS, not the kernel which is GPL), not some wacko bullshit trickery with GPL and a CLA

With regard to your contention that “work published with GPL should not become closed source,” according to the Free Software Foundation's Licensing and Compliance Lab, “[i]f you are the copyright holder for the code, you can release it under various different non-exclusive licenses at various times.” That is, the copyright holder can release a program under the GNU GPL, and can also use that same code in non-free programs.

Are you suggesting a change to current policy?

In any case, although Canonical accepts contributions to the Mir project, Canonical is by no means dependent on such contributions. They could easily go it alone, in which case Canonical would be the sole copyright holder. But then, Canonical would be accused of not playing well with others.

By accepting contributions under a their CLA, Canonical hasn’t “stolen” anything. Those who contribute under Canonical’s CLA do so willingly. The CLA is unambiguous, and no “wacko bullshit trickery” is involved.

You wrote that “If [Mir’s] original license was Apache or some other permissive then it would be totally fine as the writing was on the wall.” Knowing that, according to the Free Software Foundation's Licensing and Compliance Lab, the copyright holder can release a program under the GNU GPL, and can also use that same code in non-free programs, would “it would be totally fine” with you if Canonical continued Mir under GPL, but refused to accept any outside contributions for Mir?

Quite frankly, given a choice between: (a) Mir being released under a permissive license (such as MIT) from the start; or (b) Mir being released under GPL, but with Canonical retaining the right to release an improved non-GPL version of Mir later (which right a copyright holder has by default and is permitted to do), I don’t get why you would prefer option (a). You seem to prefer GPL over permissive licenses. Say what you will about Canonical’s approach, but you can’t deny that option (b) results in the release under a massive project under GPL that will forever “remain out with the same [GPL] license” while option (a) doesn’t.

As for why Google didn’t pursue something like option (b) with Android, I suspect that Google has its own competitive (and perhaps legal) reasons. But in any case, the choice was Google’s to make.

And yes, I realize that option (b) may afford Canonical a potential advantage over potential competitors who would like to use Mir code in competing non-free programs. However, like I wrote before, while contributions to Mir are welcomed, Canonical's development of Mir is by no means dependent on contributions. Mir is Canonical’s baby, and I wish Canonical that best of luck and much success.

[starshipeleven]

Great! Problem solved!! Now that Canonical's CLA is a non-issue, we can all move on and agree never to discuss Canonical CLA again. Yippee!!

The fact that contributors avoided CLA projects does not make it any less of an issue.

It just backfired, and now it is a Canonical issue.

In all seriousness, although I don’t have a crystal ball, I don’t foresee Canonical ever attempting to exercise their option to relicense Mir under a license other than GPL.

The issue here is that they have the choice to do so, but only them. GPL license is supposed to NOT ALLOW that. Which makes the whole thing weird.

Really there is nothing wrong with choosing an opensource permissive license, but no, they wanted to keep the right of making it closed for themselves only.

CLA includes several other notable provisions, including a provision that seeks to prevent a contributor from later suing the project for infringing that contributor's patent(s) covering the contribution

Yeah, patent clauses are totally unheard of in other permissive licenses.... no wait.

Yet, no one has been able to “pull tricks” or “pull an EEE” with the X Window System in any way that has resulted in “causing massive havok and forcing people to do their bidding or face massive issues in either maintaining themselves the last opensource stack or migrate to something else.”

The fact that in some cases it did not happen does not mean that it cannot happen. see also below.

Assuming the GPL version of Mir ever becomes “very common everywhere” and "widespread" on Linux desktops (which even you doubt will happen), what leads you to think that Canonical could “pull an EEE” on the community?

Linux desktops don't generate revenue so it would not matter to even try it.
The target of an EEE are either company-grade distros (that can and do ship their own proprietary stuff) or similar, and rely on the money from their customers to stay profitable.
Which is why RedHat especially was so afraid of CLA.

Still, that's not how licenses work.
Licenses are supposed to make sure something does not happen as they specifically rule out things.

Otherwise a solemn oath over the bible from Canonical would suffice.

according to the Free Software Foundation's Licensing and Compliance Lab, “[i]f you are the copyright holder for the code, you can release it under various different non-exclusive licenses at various times.” That is, the copyright holder can release a program under the GNU GPL, and can also use that same code in non-free programs.

Newsflash, the copyright holder can always change the license of his own property, or make a copy of the same work with a different license. They don't need anyone's permission to do so, because that stuff is their own.

Are you suggesting a change to current policy?

DId canonical write the CLA for themselves or for contributors? Because here the issue is the contributions.

By accepting contributions under a their CLA, Canonical hasn’t “stolen” anything.

Inappropriate word, ok, but I wanted to convey that they are asking people to give them the same privilege of the author, while not allowing this to anyone else.

Which is more like a very big gift, and not so common thing to ask in opensource. Both the copyleft camp and permissive camp don't usually like making exceptions, the same rules apply to all contributors/users of the project.

would “it would be totally fine” with you if Canonical continued Mir under GPL, but refused to accept any outside contributions for Mir?

Yeah it would be fine as the issue here is the contributions, not what Canonical writes, because since they are the copyright owner they can re-license it however they want, whenever they want.

It would make 0 sense though, worst license for an embedded product, while no contributions (which is the point of opensource licenses).

Quite frankly, given a choice between: (a) Mir being released under a permissive license (such as MIT) from the start; or (b) Mir being released under GPL, but with Canonical retaining the right to release an improved non-GPL version of Mir later (which right a copyright holder has by default and is permitted to do), I don’t get why you would prefer option (a)

I never said I would prefer option A.
Also, the bolded part is misleading. While they can do so for their code, the issue here is contributions, Canonical isn't the copyright holder of the contributed code.
So B isn't an option without the CLA that grants them the ability to do so.

As for why Google didn’t pursue something like option (b) with Android, I suspect that Google has its own competitive (and perhaps legal) reasons. But in any case, the choice was Google’s to make.

Google chose Apache for the same reason Canonical added CLA, the embedded market they are targeting is very afraid of GPL.

Canonical's development of Mir is by no means dependent on contributions.

Well, that's open to debate. If they had decent contributions they would have a working product they can sell by now, not a half-finished one that is barely ok for simple embedded projects (where you can get away with much much less than a full-blown display server).

The whole point of being opensource is sharing the load, if you are not leveraging that to your advantage, you're using opensource wrong (in the sense of not getting any benefit from it).

And yeah, Canonical has a long tradition of using opensource wrong.