6

The website of a library licensed under AGPL-3.0 states the following (emphasis mine):

AGPL-3.0 License

This OSI-approved license is designed with students and enthusiasts in mind, championing open collaboration and the free exchange of knowledge.

It requires that all software and AI models under its banner be open-sourced. More importantly, any larger project or software incorporating an AGPL-3.0 component must adopt the same open-source stance.

This policy not only ensures unmatched transparency but also reinforces the bedrock principles of collaborative innovation in the tech world. If you engage with AGPL-3.0 licensed tools or AI models, be ready to open-source your entire endeavor.

Based on my understanding, AGPL-3.0 only requires you to provide anyone who interacts with the derivative software access to the source code. This wouldn't necessarily entail open-sourcing the code as I could only disclose it to the users of my software. However, the above paragraph seems to contradict that notion and says the derivative will also have to be open-sourced, not simply made available to the recipient of the derivative software. Open-sourcing it implies that anyone should have access to the source code of the derivative, whether they are interacting with the derivative software or not, which is different from providing access to the source code to people who interact with it.

Is the paragraph misconstruing the AGPL-3.0 license or is there such a requirement due to the original code being public?

Improve this question
5
  • 3
    FWIW, that ‘website’ seems to be <ultralytics.com/license>.
    – Dmitry Alexandrov
    Commented Nov 4 at 5:48
  • Could I beg elements of the community to stop commenting here telling people that they've edited the question, and saying why? This site's a wiki; anyone can see the edit history of the question, so nobody needs to be told that. As for motivations, that's really not the purpose of comments.
    – MadHatter ♦
    Commented Nov 4 at 11:42
  • @MadHatter, to beg? Why to beg, when you can just (ab)use your moderatorʼs privileges and any comment, that references your lovely self in way you dislike, will disappear without a trace or even notification? Then you can always invent a plausible pretext, like ‘nobody needs to be told that’.
    – Dmitry Alexandrov
    Commented Nov 5 at 10:29
  • @MadHatter, btw, I have no idea, why have you so much hated to have a very good excuse (the initial lack of context provided) for not been able to see, that it was not OP, who had to be reeducated to use the ‘correct’ meaning of words ‘open source’, but Ultralitics who played on customersʼ fears to discourage them from using the AGPL-covered version.
    – Dmitry Alexandrov
    Commented Nov 5 at 10:30
  • 1
    @DmitryAlexandrov Part of Stack Exchange's method to Q&A is that comments are ephemeral -- if you have substantial things to say about the question, you can either embed them within the answer (as your edit has done) or include them in an answer. For example, the idea that the OP's confusion stems from misinformation from Ultralitics is substantial idea that deserves a place in answering the question, so ought to be in an answer, not a comment. By contrast, justifying or deliberating edits is ephemeral; this edit stands, so the comment can be removed.
    – apsillers ♦
    Commented Nov 5 at 12:18

4 Answers 4

Reset to default
13

You seem to have set this up as a dichotomy - a choice between source-only-to-users and fully-open-source - but it's not.

You are right that the GPL requires you to provide source code only to recipients of the binary, and AGPL extends the benefit of that requirement to individuals interacting with the binary over a network. But what you may be missing is that when you provide these people with the source they may then themselves use it under the terms of [A]GPL.

So:

  • no, you don't have the give the source to anyone other than a recipient of the binary, and in the case of AGPL anyone interacting with the binary over a network, but
  • the people to whom you give it may in turn give it to anyone at all, and they may modify it before doing so, if they so choose.

This isn't self-contradictory. Free software has always been about the rights of the users of the software: not the developers, nor the copyright holders, nor the rest of the population of the planet. If you use the software, you have the rights to copy it, to modify it (and the power to do that via the source code), and the right to distribute copies of either verbatim or modified software, the latter term including things which are classified as derivative works under copyright law.

Improve this answer
5
  • "the people to whom you give it may in turn give it to anyone at all, and they may modify it before doing so, if they so choose." - this is fine by me. But I'm more concerned on whether I have violated the license if I built a project using that library and handed over the source code and the software to the client, but didn't make it open-source myself, as in release it publicly.
    – Crimson
    Commented Nov 4 at 8:06
  • 6
    That's not what open source means! As my first bullet point is pretty clear, I think, free software does not require you to give the source to everyone. You are required to give it to anyone who gets the binary (plus, for AGPL, anyone who interacts with the software over a network). Moreover, that conveyance must be done under [A]GPL. That is what open source (and equally, free software) means.
    – MadHatter ♦
    Commented Nov 4 at 8:10
  • @MadHatter wrote: “If anything, the problem is that the OP misunderstands the term "open-sourcing”. Why are lecturing the OP, instead of answering his question straight? He definitely has enough understanding of the matter to correctly spot a language trick, employed by the vendor to lure him into buying nonfree version.
    – Dmitry Alexandrov
    Commented Nov 4 at 9:16
  • 1
    @MadHatter "That's not what open source means." - I guess my confusion is that I am associating something as being "open source" as also being public (authoritative answer on this is really hard to find; even Wikipedia mentions "public" several times). So something can be considered "open source" even if it is confined between certain parties.
    – Crimson
    Commented Nov 5 at 0:22
  • 3
    @Crimson as may be becoming clear from some of the (quite heated) discussions being had on this page, the term "open source" is just as ambiguous, and subject to misinterpretation and marketing, as "free software" ever was. The term has no statutory protection, so people can quite legitimately use it to mean whatever they like; and they do. You can't find an authoritative statement, because there is no authority. Here, by community choice, we use the OSI's definition, and that definitely doesn't mandate publication of the code.
    – MadHatter ♦
    Commented Nov 5 at 7:02
6

You are right that you only need to pass a licensed copy of the sources to your users. That might be closer to making it public to everyone right away yourself than you think, though:

Assume, I'm one of your users, who interact with your AGPL-software. Thus now I have the right to get the sources from you under AGPL. The AGPL license is an open source license and grants me the right to do whatever with the software, including making it publicly available under the AGPL to anyone and everyone.

Given this, the quoted passage is correct IMHO (though technically making the source available is only required to those who are your users)

Improve this answer
2

The AGPLv3 is one of the strongest copyleft licenses that exist.

The AGPLv3 license (without additional permissions of Section 7) applies to all libraries and software statically and dynamically linked to AGPLv3-licensed software, except System Libraries. It applies to the source code, build scripts and signing keys (for Secure Boot).

This means that if your code is linked to AGPLv3-licensed software, then this whole combination is subject to the conditions of AGPLv3.

Anyone receiving binaries of this combination and/or anyone interacting with this combination of software through a computer network like the Internet, has the right to obtain the full source code (the Corresponding Source) of the combination, as AGPLv3 downstream licensees.

As AGPLv3 downstream licensees, they can then use, modify, distribute (like putting it in a public GitHub repo), make it interactive over a computer network and/or sell this combination, again subject to the conditions of the AGPLv3.

So, in effect, your code too becomes licensed under AGPLv3. It becomes open source software.

The only chance your code can escape this is if it is not linked to AGPLv3-licensed software and it communicates with AGPLv3-licensed software solely through inter-process communication methods like TCP.

Improve this answer
1
  • 2
    I think your answer doesn't sufficiently address the crucial contrast between privately distributing something as open source vs "open sourcing" it as in publicly making it available (quoting the question's wording).
    – ecm
    Commented Nov 3 at 16:26
-5

Does [GNU] AGPL-3.0 require open-sourcing the derivatives if the original work is open-source [= public]?

No!

That requirement would actually render it a nonfree licence, which would be ironic for a GNU licence.

Is the paragraph misconstruing the AGPL-3.0 license

Yes.

Moreover, not only that paragraph, from the very first word: ‘OSI’ this piece of text misconstrues the context, intents and of purposes behind GNU (A)GPL.

Based on my understanding, AGPL-3.0

Reading the licence itself, and its preamble in particular, you must have noticed, that it says absolutely nothing about ‘open source’.

If you go a bit further and follow the hyperlink in its footer, you will find (maybe to your surprise?), that itʼs primary author (Dr. Richard Matthew Stallman) is a very vocal opponent of Open Source.

In 2024, in English, saying ‘open source’ is often just as matter of choosing a mere synonym for the original ‘free’, that is way less ambiguous, and is free of unwanted (in some cultures, and thus in international corporate context) connotations.

But in in the eyes of their respective envisioners they were not synonymous. And your question nailed the most crucial difference!

In RMSʼs (and thus Free Software Foundation and GNU Project) eyes, the freedom to distribute software to anyone, including to no one, i. e. to enjoy it privately, is one of the four most essential ones, that any free licence must guarantee.

And of course, GNU own licences, Affero GPL v3.0 included, do that.

But the aforementioned OSI (Open Source Initiative) never took this stance.

On the contrary: they approved a couple of licences, which indeed included such a requirement — to ‘open source’ (to share) all your modifications with everyone (Reciprocal Public License v1.1, RPL v1.5).

Thankfully, nobody really cares about OSI, so your probably never heard of them. ;)

Improve this answer
10
  • 2
    I'm not saying I disagree with you, but this question's not about the age-old free/open-source debate, so your answer isn't the right place to relitigate that. I also note that on this site we use the FSF's definition of free and the OSI's definition of open-source, so for nearly all practical purposes they may be considered the same. That means your very first word is wrong, as AGPLv3 absolutely does require that derivative works be published under AGPLv3 (see AGPLv3 s5c).
    – MadHatter ♦
    Commented Nov 4 at 8:19
  • @MadHatter wrote: “for nearly all practical purposes they may be considered the same”. Yes, except the cases, when they may not, such as this one: that concerns the ‘desert island test’.
    – Dmitry Alexandrov
    Commented Nov 4 at 8:47
  • 1
    Note that the OSI definition of "open source" contains nothing that would mandate distribution, any more than the Four Freedoms do, so I don't see that the "desert island" test is relevant here. If anything, the problem is that the OP misunderstands the term "open-sourcing". But you don't have to perpetuate their misunderstanding in your answer - an answer which, if read without that misunderstanding, is simply wrong.
    – MadHatter ♦
    Commented Nov 4 at 8:52
  • @MadHatter wrote: “on this site we use ... OSI's definition of open-source... That means your very first word is wrong”. Who are those ‘we’? Crimson the OP clearly used it another sense, which means that the straight ‘No, it does not; yes, Ultralitics claims are bullshit; and you are right’ is the only correct answer to his question.
    – Dmitry Alexandrov
    Commented Nov 4 at 9:00
  • 2
    I never said the thing you quote me as saying, so I'm not sure that's the most-sincere-looking de-escalation in the history of internet interactions.
    – MadHatter ♦
    Commented Nov 4 at 10:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.