Marc Haber <[email protected]> writes: > On Tue, Aug 11, 2026 at 04:10:00PM +0200, Gard Spreemann wrote: >>In order to safeguard the project's mutual understanding, trust and >>respect, we disallow the output of generative AI as direct contributions >>to Debian. > > This wording leaves HUGE possibilities of interpretation and will therefore > cause a lot of flamewars.
I don't think this is any more true for this option than for the others, sadly. > Claude is not a DD, it therefore cannot contribute directly to Debian. When I > say "Claude, create a Debian package doing foo and set myself as a > Maintainer", > is that a direct contribution? Assuming that the software itself already exists upstream and that you are talking about its Debian packaging: Yes, this is in fact the archetype of the kind of contribution I had in mind. Assuming that you start from scratch and write the software itself using Claude, while doing the Debian packaging part as a human: This is indeed a loophole – the software itself is now an indirect contribution by virtue of being upstream. I would find it strange to attempt to forbid a DD from writing non-Debian software in whatever way they want. And strange to forbid *just* DD-written upstream software from entering Debian. So the loophole must remain. I do think people understand that exploiting this loophole would go against the spirit, if not the letter, of the GR if this option were adopted. > Does the direct contribtion become indirect when I fix the Lintian errors that > Claude causes myself? No. > Does it stay a direct contribution when I just prompt Claude to fix > the Lintian errors? I think we're talking past each other. And that is of course a sign that my language in the proposal isn't clear enough. The distinction between "direct" and "indirect" has nothing to do with an LLM being used. Those are words I use to try to make it clear that the proposal doesn't say anything about AI-generated code ending up in Debian "indirectly" by for example being part of an upstream project (incidentally: for me, this is just about pragmatism, not a fundamental difference). > Does that become indirect again when I review the diff? You might be interested in Theodore Tso's objections to the proposal. My answer to this is there are clearly situations where the diff of a fix made by a human and that of one made by an AI are indistinguishable. While the proposal does not directly address this, I do not think that is where we should focus our energy. Assume good faith, and thus assume the diff represents a human fix (even if it doesn't). A mediocre analogy might be a snippet of assembly. If that assembly is the output of a compiler, we would in Debian prefer to have the source code the compiler ingested as the actual source code. If that snippet of assembly is so tiny that a human writing manually and a compiler compiling must essentially produce the same thing, then I don't think it's worth arguing over how that snippet was actually produced. Best, Gard
signature.asc
Description: PGP signature
