James Le Cuirot <[email protected]> writes:

> On Tue, 2026-03-10 at 17:16 +0000, Filip Kobierski wrote:
>> Hi Michał,
>> 
>> The issue you described is real and widespread.
>> 
>> Maybe one could flag slop packages with a LICENSE variable that is not 
>> accepted by default?
>> That would allow users to still have the final say in what can run on their 
>> Gentoo systems but would be aware that AI-SLOP license is suboptimal.
>> Then I imagine the problem would be in marking packages as such...
>> Also the name of the "LICENSE"; AI-SLOP seems in-line with Gentoo's 
>> approach, alas I in my opinion is unprofessional. Plain AI does not really 
>> sound discouraging. Naming is a secondary issue though.
>> 
>> What do you think about that?
>
> If a maintainer dislikes something about a project strongly enough to stick
> this kind of warning label on it, then I imagine they won't want to maintain
> it anyway. If it's a dependency of something they care about, then they could
> pass it onto someone else, although they may still be unsatisfied with that. I
> don't know what else to suggest. Forking is not the answer, as mgorny has
> written elsewhere.
>
> If they're otherwise willing to hold their nose, I think a simplistic label is
> not going to be sufficient. Different projects are doing different things, and
> it's not fair to tar them all with the same brush. In addition, everyone has
> different opinions and tolerances around this. I personally think what chardet
> did was irresponsible, but I don't have an issue with what KeepassXC are
> doing. I therefore think the best thing to do is have a free text field in
> metadata.xml where a maintainer can fully express any concerns along with
> links to evidence, much like what mgorny has done here. That way, users can
> make their own fully informed decisions.

OK, but what do we do about the dev-python/chardet case? How do we
signal to people that they shouldn't bump to it and shouldn't depend on
>=7 (the bad version)?

We can rely on people "just knowing" for chardet because it's maintained
by @python, but what do we do for maintainer-needed packages say in this
state?

Your stance works (somewhat) for leaf packages, to some extent. It
doesn't work for two other cases:
1) where users wish to avoid such packages and have a standard way of
finding out about it (let's ignore KeepassXC for now, as I think it's
contestable enough);
2) other packages depending on it and needing to lobby upstreams or
patch downstream to avoid >= deps on something infected

I'll write this up separately.

Attachment: signature.asc
Description: PGP signature

Reply via email to