[gdal-dev] Tightening AI based submission policy

[Greg Troxel]

Daniel Baston via gdal-dev <gdal-dev at lists.osgeo.org> writes:

> As one of the maintainers, I am comfortable with the current policy. While
> I wish we did not live in a world of low-quality AI-assisted pull requests,
> they are not at the top of my personal list of burdens. If a PR appears to
> be slop, I simply do not spend time reviewing it. We clearly have had
> submissions that aren't compatible with the policy, and that will be the
> case with whatever policy we choose.

I think I understand your comments but I'm having trouble with "and thus
it's ok".

In a different context, pkgsrc, there are N people who can commit, and
when a new upstream release comes out any of them can update to it.
Depending on the upstream bits, that might or might not be a good idea.
Since the bits end up based on the one that acts, vs the N-1 who think
"upstream .0 release is too flaky; we should wait", mechanisms are
needed to ensure consensus.

Merging PRs is more or less like that, and that leads to people who
review thinking they need to pay attention to PRs that someone else
might merge and say no, this is not ok.  Just deciding not to review
them doesn't reliably lead to the right outcome.

Given the current policy, I would expect LLM-containing PRs to be simply
closed when a judgement is made that that's what they are.  But in proj
(which is not gdal I know), I see labels saying they are LLM and
violating policy -- but still open.  A quick glance at gdal repo shows a
similar situation less the explcit label.

Of course people (bots?) will violate any policy.  The question is how
to have a policy so that LLM-containing submissions that violate the
policy will be quickly closed.  Not labeled nonconforming and left open,
but closed.  People who value community acting in good faith will be
reluctant to do so -- so the policy/group process needs to make it easy.

When I see likely-LLM and it's not disclosed, especially on mailing
lists, I just ask.  The responses have been mixed "yes it's LLM" and
just silence.  This is still work, but I don't feel at all bad saying
"this doesn't belong and I'm not going to read it" after an admission of
LLM, and I don't mind hitting close if there is no response.