Re: LLM Commit in upstream `master`?!

Eli Zaretskii [26/Jul  7:56am +03] wrote:
> I don't see any problem with having the commit log message generated
> by a program.  Those messages are not covered by copyright, and are
> quite short, so I see no problem in using LLM for that.  Why do you
> see a problem, and what is that based on?  The GNU policy is only
> talking about code, AFAIU.

I think you are probably right that the contents of the commit message,
in this case, is not substantive enough to cause issues.
At the same time, I think we should take the maximally cautious approach
until and unless we get clarity from the FSF about the policy.

> We should remove the Co-Authered-By heading, but that can be done by
> editing the generated ChangeLog.
> [...]
> We could do that, but forced commits are trouble for everyone, so we
> should think hard whether this is justified.  We have more grave
> mistakes in commit messages, like the missing
> Copyright-paperwork-exempt thingies (which _are_ legally significant),
> but we never even bother fixing them.  If the only problem is to cope
> with people who ask questions about that, I'd say "bring them on".

I am convinced it's justified in this case.  We don't want people doing
data mining or searches on the Git history of our project to think that
the body of this commit was written by an LLM.  We have no sensible way
of recording the truth about that commit.  The fact that you and I know
it's probably okay is not good enough in this case.

-- 
Sean Whitton