Hacker News Reader

I am in fact the prime mover behind License Zero (https://licensezero.com), but I am by no means alone in it. Parity and Prosperity would not be what they are without enormous amounts of feedback and suggestions from users and interested hackers. Some of those folks agreed to be listed here: https://licensezero.com/thanks

There may be more copyleft licenses than you think. Blue Oak Council, in which I'm also involved, recently published this resource: https://blueoakcouncil.org/copyleft#copyleft-families

That being said, you're absolutely right that we've seen more permissive licenses than copyleft license, by simple count. Compare: https://blueoakcouncil.org/list

> limits commercial use to a short trial period of 32 days. When a commercial user’s trial runs out, they need to buy a private license or stop using your software.

So it's clearly not Free Software because it doesn't respect freedom 0. Hardly a surprise that Bradley Kuhn and the FSF object to it being classified as such.

This is hardly a failure of "free software", just the author wants to try to subvert the definition which the FSF has long established. Maybe the author should try calling it something completely different (I suggest "trialware") and then the FSF wouldn't be bothered.

(https://www.gnu.org/philosophy/free-sw.en.html)

I’ve never presented Prosperity, the noncommercial form, as free or open source. The debate was about Parity, the strong copyleft form.

Why not simply use the GPL?

The GPL has always been criticized for its definition of distribution, which allows companies to host SaaS offerings of GPL software without having to make their modifications available.

Hosting a GPL-licensed server and receiving network activity does not constitute distribution, which is why the AGPL exists.

> However, I always found it a bit strange that there were dozens of variants of the MIT and BSD license out there, but aside from the GPL and MPL no other popular license really took up the mantle of copyleft, aside from perhaps the old Sleepycat license.

Fundamentally, copyleft can only be instantiated once. No two distinct copyleft licenses as we know them can be compatible with each other without jumping through carefully-placed hoops[1] because the combination necessarily imposes stricter distribution conditions (there would be no point otherwise), which is expressly what each copyleft license seeks to prohibit per se.

I think it's only a slight exaggeration to say that the perceived problem of license proliferation, and the ecosystem of license gatekeeping that sprung up around it, owes itself wholly to the design of the GPL in this respect.

It didn't last. The GPLv2/3 split haunts us to this day.

1: e.g. https://www.mozilla.org/en-US/MPL/2.0/FAQ/#mpl-and-lgpl

> Fundamentally, copyleft can only be instantiated once.

That's true of GPL-style copyleft, because it was designed quite intentionally to be a network-effect-leaden singleton. We say "GPL-compatible", but it's more accurate to say "GPL-submissive". The difference was felt very directly by the Eclipse folks in the run up to EPLv2.

But there's nothing about share-alike terms that require that stricture. The Parity license that I mention in the blog post takes a different tack. Have a look at its rule about how to contribute back: https://paritylicense.com/versions/7.0.0.html#contribute

A developer building on Parity code could choose to license their work on weaker copyleft terms.

> But there's nothing about share-alike terms that require that stricture. The Parity license that I mention in the blog post takes a different tack.

But they do. "or another license that allows everything this license does" sounds less restrictive at a glance, in that on the surface it permits additions to be separately licensed, but its effect is the same: no license with more stipulations is acceptable, and by symmetry any less restrictive share-alike license with similar wording cannot accept combination with this one.

Additions to code under your license may be published under weaker terms. But they cannot be weaker copyleft terms. I can think of a few edge cases involving use of independent portions of the new work under only the weaker license, but it doesn't achieve a likely goal of someone choosing a copyleft license: to eliminate confusion about whether it's permissible to merge downstream improvements.

> But they cannot be weaker copyleft terms.

Why not? What about, say, the API Copyleft License, which reads very much like Parity, but permits building closed applications? https://apicopyleft.com/

That license allows everything Parity does, plus one more: creating closed applications.

This so-called API Copyleft license is too weak to usefully consider a copyleft license.

Do you think the same of MPL? https://spdx.org/licenses/MPL-2.0.html

Taken on its own merits, I would indeed consider MPL 2.0 inadequate as a general-purpose copyleft license. Formally, you could argue that it's even weaker than the API Copyleft terms. As such, it's hard to recommend MPL for a project standing alone.

It understands, however, its position as a compromise which makes the underlying goal of its approach to copyleft untenable. Despite that, tries to preserve the ability to incorporate downstream changes through one degree of separation. The difference in expectations is palpable. The implied flow for publishing derivatives of MPL work is to distribute also under MPL, with the option of separately licensing non-Covered Software presented as a concession to tightly define the scope of the license. The implied flow for publishing derivatives of API Copyright Licensed work seems to involve carefully considering how much you can get away with under the listed exceptions, with the option to not share alike presented front and center.

The distinction is essentially irrelevant to anyone approaching licensing from a legalistic perspective, bet defaults matter for everyone else.

The use case for MPL and SSPL was broadly the same: selectively apply permissivity for the primary build-on use case (browser plugins, applications) to stoke adoption and copyleft to other use cases (forks, hosted offerings) as a bulwark against competitors (Microsoft, AWS). In drafting and presentation, I thought SSPL failed to communicate that heritage. So I wrote what became the API Copyleft License.

It is always useful to fall back to basic when ever people try to redefine copyleft or GPL. A copyright license is a list of complex permissions. Nothing more, nothing less.

If you have a two set of permissions, compatibility is the logical conclusions of finding intersection and see which actions are permitted in both set. Set theory do not have submissive and dominant sets. The intersection is always the elements which are in both A and B.

The parity license says "Make sure every part of the source code is available under this license or another license that allows everything this license does". It practically spells out that permissions in the intersection that lies between parity and an other license is acceptable to use when combining parity with an other license. It is true. It also true for any other copyright license, including GPL.

Set theory doesn’t, but this isn’t pure set theory, it’s a social process (perhaps it’s game theory, though I don’t think you have to really use math to get the point ). The GPL came first, so it effectively forces others to submit or to go it alone.

Copyright is the entity that came first. A copyright license is the object you can use to claim you have permission to do something which copyright declares illegal by law.

We can put gpl first when trying to determine the intersection of all the set of permissions when dealing with multiple licenses, but in the end the fundamental question to be answered is: Do you have a set of permissions to do something which copyright makes illegal. Yes/No. In the case of multiple copyrighted work you need permission for all that which copyright limits in those cases.

If we want to look at it from a game theory perspective it we would ask who the agents are, their goals, the environment and the resources. Here a narrow intersection can be a cost, or it can be irrelevant depending on context. The biggest permission is always to write it yourself which make the limitation of copyright a non-issue for the author. Naturally that is not the only factor, nor the only agent, so it all depend.

You’re technically correct, and I think set theory is a very productive metaphor for the analysis. But that analysis with GPL hasn’t worked out so cleanly, however, due to its complexity. Even GPLv2 is a relatively long license. It attempts to spell out copyleft in relatively low-level terms, legally speaking. Add on additional cleverness, like anti-TiVo and liberty-or-death, and it becomes very hard to say where the boundaries are, or to compare them to less idiosyncratic licenses.

It is fair to say that a complex description of permissions is harder to define and compare than a simple description. GPLv3 attempts to be international interpreted across different copyright laws and to a degree different languages, which makes the job even harder.

But it also fair to point out that the basic permissions are fairly simple. Give the recipient source code of the whole work, put it under the same license, don't do something which a judge would deem as circumventing the authors intention of giving you those permissions.

The above simplified version of copyleft would likely work fine, although I suspect every lawyer on the globe would hate it.

> The above simplified version of copyleft would likely work fine, although I suspect every lawyer on the globe would hate it.

I’m familiar with several who do not!

Isn't commercial license revokable?