GPL upgrades via section 14 proxy delegation

2026-03-068:4110955runxiyu.org

I am not a lawyer. This is provided solely for the purposes of general information and does not constitute legal advice, guidance, or counsel. No attorney-client relationship is established by the…

I am not a lawyer. This is provided solely for the purposes of general information and does not constitute legal advice, guidance, or counsel. No attorney-client relationship is established by the provision of this information, and no reliance should be placed on this information in lieu of seeking professional legal advice. All information is provided “as is”, without warranty of any kind, either express or implied.

Problem

Let’s say that you wrote a piece of software, and publish it under the GNU General Public License, version 3.0, or the GNU Affero General Public License, version 3.0 (for the sake of simplicity let’s just call them GPL because the networking clause is irrelevant).

You would need to choose between two variants:

  • GPL-3.0-only: Even when the Free Software Foundation publishes a newer version, your project will stay on GPL version 3.0. If you are the sole copyright holder, you could update to another version of the GPL (or, for that matter, any license) at your choice. If there are other copyright holders (such as contributors1), you cannot unilaterally change the license, and would have to ask for permission from each copyright holder (or remove their code).

  • GPL-3.0-or-later: When the Free Software Foundation publishes new versions of licenses, anyone could choose to use your program under the terms of the new license.

I find neither approach to be ideal. It is often impossible to gain consensus of all copyright holders since some may be unreachable. I also wouldn’t trust the FSF with the power to unilaterally change the license of my program.

Solution

Notice that in section 14,

If the Program specifies that a proxy can decide which future versions of the GNU Affero General Public License can be used, that proxy’s public statement of acceptance of a version permanently authorizes you to choose that version for the Program.

So, in my projects, I simply write something like this:

This project is licensed under the GNU Affero General Public License, Version 3.0 only.

Pursuant to Section 14 of the GNU Affero General Public License, Version 3.0, Runxi Yu is hereby designated as the proxy who is authorized to issue a public statement accepting any future version of the GNU Affero General Public License for use with this Program.

Therefore, notwithstanding the specification that this Program is licensed under the GNU Affero General Public License, Version 3.0 only, a public acceptance by the Designated Proxy of any subsequent version of the GNU Affero General Public License shall permanently authorize the use of that accepted version for this Program.

For the purposes of the Developer Certificate of Origin, the “open source license” refers to the GNU Affero General Public License, Version 3.0, with the above proxy designation pursuant to Section 14.

(… Standard DCO stuff …)

It’s patently clear2 that the license allows this, and it surprises me that this is rarely brought up in debates about GPL-3.0-only and GPL-3.0-or-later.


Read the original article

Comments

  • By ognarb 2026-03-069:30

    We do that in KDE too, where the decision to update to a possible gpl4 is decided by a vote of the KDE e.v. (the legal non profit organization behind the project) membership.

    https://invent.kde.org/office/marknote/-/blob/master/LICENSE...

  • By shevy-java 2026-03-068:565 reply

    > I find neither approach to be ideal. It is often impossible to gain consensus of all copyright holders since some may be unreachable.

    Well, licences are not universal wonder tools. They have restrictions about their use cases. But, narrowing this down solely to "GPL xyz" versus "GPL xyz - or later fancypants", I always found the variant WITHOUT the "or later" to be better. It simply adds more complexity when a licence can willy-nilly be changed, at a later time, when a change happens. I understand the use case for the "or later" part, as the GPL is very strict as well as an ideological tool against abuse from corporations (let's be honest here; and I think the GPL is a good licence, despite this too), but even then I find it better to stick to the simpler variants. It is one reason why I may use GPLv2. I also use MIT/BSD when I essentially don't care much. I don't think I have had a use case for GPLv3; and not for "or later" either. LGPL is also fine.

    > It’s patently clear that the license allows this, and it surprises me that this is rarely brought up in debates about GPL-3.0-only and GPL-3.0-or-later.

    I was unaware that a proxy can be designated upfront; so that's another complexity with regards to the "or later" part. What can proxies do? I dislike the "or later" clause; it really just makes this way more complicated than it should be.

    • By zvr 2026-03-0615:58

      The main advantage for using "or later" is not really to be OK when a new version of the license is published, as this happens rarely.

      What you gain is the possibility of combining this code with any other code that is under a later version of the license. If there is code X under GPL-2.0-only and code Y under GPL-3.0-only, these cannot be combined, since each license declares that any derivative work has to be under the same license. If code X were under GPL-2.0-or-later, the combination would be compliant.

    • By weinzierl 2026-03-069:15

      "It is often impossible to gain consensus of all copyright holders since some may be unreachable."

      How one feels about that is a matter of where one stands. The GPL first and foremost protects the interests of software users. Not developers. Not companies.

      In that regard, the above should be seen as a feature, not a bug. I believe it is the most effective way to protect the user from being locked-in.

    • By RobotToaster 2026-03-069:385 reply

      With the "or later" version it's a concern that in the future someone nefarious could gain control of the FSF, and publish a GPL removing most of the copyleft provisions.

      On the other hand, if Linux had used the "or later" version it could have helped prevent TiVoization.

      • By pabs3 2026-03-0611:57

        According to Conservancy; Tivo didn't do "Tivoization", the GPLv3 doesn't prevent what Tivo actually did, and both GPLv2/GPLv3 prevent "Tivoization".

        https://sfconservancy.org/blog/2021/mar/25/install-gplv2/ https://sfconservancy.org/blog/2021/jul/23/tivoization-and-t... https://events19.linuxfoundation.org/wp-content/uploads/2017...

      • By bonoboTP 2026-03-0610:151 reply

        No because tivo could take it under the gpl2. It's not an auto upgrade. The new version is optional.

        • By gzread 2026-03-0613:02

          New distros and modules could be v3-or-later.

      • By gzread 2026-03-0613:023 reply

        Linus now has come to support Tivoization. I presume this has something to do with where his salary comes from.

        • By jmalicki 2026-03-0621:31

          Linus was a little liberal about the restrictions of software freedom (boy is that an awkward phrase) even early on - e.g. his general acceptance of "binary blobs" in the kernel and such for things like NVidia kernel drivers, to the chagrin of much harder-core free software people.

        • By samtheprogram 2026-03-0615:43

          Linus never cared about that use case of the GPL. He cared about the source code sharing.

        • By kube-system 2026-03-0619:472 reply

          Anti-Tivoization is a pretty radical idea that restricts the rights of hardware developers for the benefit of software developers. Linus doesn't really care about strong-arming hardware developers the way RMS does. He just cares about the software.

          • By hn_acker 2026-03-070:441 reply

            > Anti-Tivoization is a pretty radical idea that restricts the rights of hardware developers for the benefit of software developers.

            How does anti-tivoization restrict the rights of hardware developers, considering that hardware developers can choose not to pre-install anti-tivoization-licensed/contracted software? Is it anti-tivoization that restricts the rights of hardware developers, or does copyright law do that?

            • By kube-system 2026-03-0715:401 reply

              You can also choose to not buy a TiVo?

              Imagine if the situation was the other way around and a CPU came with a hardware license agreement prohibiting any software that ran on it from validating hardware attributes, using encryption to communicate only with certain other pieces of hardware, or for that matter, any restrictions at all. All in the name of enabling hardware developers to tinker with their hardware without those pesky software developers getting in the way with their pesky encryption.

              > Is it anti-tivoization that restricts the rights of hardware developers, or does copyright law do that?

              In the way that all copyleft enforcement requires copyright, then yes… but what does that have to do with anything?

              I think Linus was spot on about the FSFs scope creep when he said:

              “The kernel license covers the kernel. It does not cover boot loaders and hardware, and as far as I'm concerned, people who make their own hardware can design them any which way they want.”

              • By hn_acker 2026-03-0723:01

                > You can also choose to not buy a TiVo?

                If TiVo hypothetically were to ship a device running an operating system built on an anti-tivoization copyleft-licensed kernel, and if I were buy the device at TiVo's asking price, the very terms of the software license would dictate that the device hardware let me install a FOSS operating system. I as the buyer would not be restricting TiVo's hardware developer rights, and I don't see how the anti-tivoization copyleft license on the operating system TiVo would have chosen to install would meaningfully restrict TiVo's hardware developer rights.

                (TiVo's hardware did not actually prevent people from installing a modified operating system [1]. My previous paragraph applies just as well to both Richard Stallman's definition of anti-tivoization and my long-lived misunderstanding of "anti-tivoization" corrected by Bradley Kuhn.)

                >> Is it anti-tivoization that restricts the rights of hardware developers, or does copyright law do that?

                > In the way that all copyleft enforcement requires copyright, then yes… but what does that have to do with anything?

                What rights of hardware developers does anti-tivoization restrict? For example, do you believe in a moral right (or a strong moral privilege) for hardware developers to install any software of their choosing on their own hardware? Most hardware-agnostic software copyright licenses and the lack of a copyright license restrict such a right/privilege. Porting most proprietary software to your own custom hardware would violate copyright (unless your port is a clean-room rewrite) because copying most proprietary software to anywhere including an instance of market-standard hardware would violate copyright. A FOSS license with an anti-tivoization clause does not prevent a hardware developer from installing or modifying covered software: the license conditions do not trigger until the software is distributed (as with the GPLv3) or a modified version of the software is run over a network (as with the AGPLv3).

                > Imagine if the situation was the other way around and a CPU came with a hardware license agreement prohibiting any software that ran on it from validating hardware attributes, using encryption to communicate only with certain other pieces of hardware, or for that matter, any restrictions at all.

                The concept that software "runs on" hardware and not the other way around makes a massive difference in how I believe hardware should be able to restrict software vs. how software should be able to restrict hardware. Letting a human validate the appropriateness (however that may be defined, including and not limited to criteria unrelated to structural integrity) of a bridge will have more moral use cases (both as a percentage of use cases and as absolute "numbers" of use cases) than letting a bridge validate the appropriateness of a human trying to cross it.

                > Imagine if the situation was the other way around and a CPU came with a hardware license agreement prohibiting any software that ran on it from validating hardware attributes

                (I would generally disapprove of both a hardware license that prohibits software from validating the hardware and a software license that prohibits hardware from validating the software.) Morally speaking, I might oppose a particular program that shuts down if the hardware has been repaired by a third-party while simultaneously supporting a particular program that shuts down if the hardware randomness module has been altered. I might oppose a particular computer that refuses to run DeCSS while simultaneously supporting a particular computer that refuses to run cryptocurrency miners.

                Does anti-tivoization prohibit hardware from validating software, or does anti-tivoization merely prohibit hardware from acting on a detection of invalid software?

                > All in the name of enabling hardware developers to tinker with their hardware without those pesky software developers getting in the way with their pesky encryption.

                How does anti-tivoization get in the way of tinkering by hardware developers, considering that (1) the hardware developer chooses which kernel and which operating system the hardware ships with and (2) a FOSS license with an anti-tivoization clause does not restrict personal use any more than an otherwise equivalent FOSS license without such a clause does? Are the hardware developer's rights restricted whenever the hardware developer can't control what a buyer does with the hardware and the software that came with it? Does anti-tivoization copyleft software restrict proprietary hardware any more than proprietary hardware restricts free software?

                [1] https://news.ycombinator.com/item?id=47273890

          • By gzread 2026-03-070:39

            The entire GPL is a radical idea that restricts the rights of software developers for the benefit of software users.

      • By hmry 2026-03-0610:19

        > if Linux had used the "or later" version it could have helped prevent TiVoization

        Only if the hardware manufacturer used a combined work of Linux and some GPLv3-only code, no? Otherwise, if Linux was GPLv2-or-later, they could just use it under GPLv2 terms and tivoize.

      • By sellmesoap 2026-03-0611:40

        GPL Vader license, pray I do not alter the deal any further.

    • By mikkupikku 2026-03-0617:291 reply

      I have long been skeptical of the "or later" clauses. I can imagine a not terribly distant future where RMS has passed away and GNU gets taken over by disinterested corporate psychos like happened to Mozilla, who then release GPL-4.0 without the copyleft, set up for industry looting of any GPL project that left in the "or later" clause.

      • By ronsor 2026-03-0618:572 reply

        I think AI will render software licenses and copyright irrelevant long before a hypothetical evil GPL-4 gets released.

        Most new (corporate-sponsored!) software is already under permissive licenses anyway.

        • By mikkupikku 2026-03-0620:53

          True.. my hope is that open weight models will progress to the point where they become viable coding agents for normies, so that even if open source dies with copyright, we will still nonetheless see a renaissance in people controlling their own computers, being able to create their own programs to solve their own problems. HyperCard on steroids, that anybody can use with no technical background. We're not there yet even for frontier models, but maybe in a few more years..

        • By anthk 2026-03-0621:101 reply

          If any AI's will be sued into oblivion from Copyright holders until the bubble collapses into itself due to LLM rot over time due to the lack of curated human input.

          • By ronsor 2026-03-0621:281 reply

            This is quite frankly not a serious scenario. Once the label "national security" gets affixed to anything, you'd better be sure it's not going away.

            Also, half of all AI development is in China. Why would China care about Western copyright holders, or rather, why would they start caring?

            • By anthk 2026-03-079:06

              Because China it's making good Wuxia movies today and they woudn't neither like being ripped off if they exported either their movies or manhua to the West.

    • By duskdozer 2026-03-0611:49

      It seems that "or later" would be putting an upper bound on the GPL restrictions? If additional restrictions are added, then users can still choose 3. If any restrictions are removed, the users can choose the later version.

  • By repelsteeltje 2026-03-0610:221 reply

    Can I (pedantically) raise an epistemic issue with:

    > Pursuant to Section 14 of the GNU Affero General Public License, Version 3.0, [Runxi Yu] is hereby designated as the proxy who is authorized to issue a public statement accepting any future version of the GNU Affero General Public License for use with this Program.

    Notice that [Runxi Yu] is an external reference, pointing to runxiyu.org.

    Wouldn't this mean that the designated proxy is (any?) future entity claiming to be Runxi Yu and substantiating that claim by demonstrating control over DNS entry for runxiyu.org could effectively upgrade the GPL licence? Or practically, if the domain registration lapses, a hacker takes control or Runxi Yu looses interest — what might happen to the license? And how would this affect any contributers?

    • By onli 2026-03-0610:461 reply

      Remember that law is not technical. This is a declaration to be interpreted. The Interpretation that a specific person with the legal name Runxi Yu is designated here is very clear, the link just a helper to identify the correct person at the time of writing.

      • By repelsteeltje 2026-03-0612:391 reply

        Thank you for pointing out this mistake. Of course, there also is nothing technically preventing anyone to ignore the GPL; the license itself is "just" some legalese.

        I do believe, though, that these kind of references (from paper into the real world) often introduce surprising gotchas. Especially when they are intended to address some future (mostly unknown) issue.

        The designated anchor point (person, technological artifact, legal entity) is itself often more likely subject to change than the thing it's trying to govern. Persons may be hit by a car, registries may expire, companies may go bankrupt. Governing laws may change. Countries may cease to exist...

        • By bombcar 2026-03-0615:22

          The LAW® has literally millennia of dealing with these kinds of things - especially with regards to physical property, the definitions of which may refer to a king of a country that hasn't existed for five hundred years. You can find all sorts of examples, look to the US southwest or Europe or any country that has been controlled by another for a time, and then stopped.

HackerNews