[bug#38846] [PATCH 4/4] DRAFT doc: Add a cooption policy for commit access.

[zimoun <zimon.toutoune@gmail.com>]

Hi,

My end-user opinion does not matter. But the transparency is good. :-)


On Tue, 7 Jan 2020 at 01:19, Brett Gilio <brettg@gnu.org> wrote:
>
> Tobias Geerinckx-Rice via Guix-patches via <guix-patches@gnu.org>
> writes:
>
> > Probably just me, but this glosses over maintainer approval just a bit
> > too deftly, and that even with 3 signed referrals commit access isn't
> > guaranteed, just extremely likely.
> >
> > Unless that will actually change, I think we should briefly mention it
> > as well.  People react worse to ‘let's try again later’ when they
> > think they've already passed.  Understandably so.

[...]

> This is definitely not just you. I felt similarly, as per a previous
> email on the matter where I suggested that it be 3 commiters and 1
> maintainer. But that process turned out to be redundant, if not
> completely superfluous by Ricardo's mention of how the process is likely
> to change in the future with a different integration model.

I am not native  English speaker so maybe I misread. From my
understanding, the proposal clarifies how the access is granted.
Currently, only the maintainers grant; from what I have observed, it
is more because of historical reasons than an explicit model. The new
integration model proposes to enforce the trust and goes to
reduce/distribute the "power" of maintainers -- which is good IMHO.
Therefore, the maintainers trust the current committers, and if 3
committers approve, why should 1 maintainer reject the applicant? It
is a chain of trust.


> Regardless, I hear your point. I also think that getting refused after
> achieving three referrals is a hard point. I think it should be
> documented clearly that the mainters have the final say.

I do not see why. If 3 referrals vouch an applicant and 1 maintainer
refuses, then there is an issue elsewhere. The chain of trust is
broken and it means that the project is not healthy.


> Additionally, and this is just a point for my part, depending on what
> kind of merit we are taking for credence in a committer making a
> referral, should we only consider committers who have worked closely
> with the person requesting commit access, or is somebody who has
> reviewed and seen their patches in passing also a viable subject?

As an end-user, I do not want to pull "bad" code; which means not GNU
compliant. And a rule of thumb usually is: when you (committer) are
annoyed to review patches because you know that you would not do
better yourself; concretely, significant contributions with no final
tweaks.


> For example, I have been asked a few times by people to push patches for
> them over IRC, but their patches were unrelated to software I use /
> would use / know how to approach (examples being GNOME). So, I kindly
> refused to push their patch citing that I do not feel comfortable in
> knowledge to understand the ramifications of their
> patches. Hypothetically, if such a person approached me in the future
> and asked for a commit access referral, since I had not worked closely
> with them what kind of weight would be referral hold?

The bottleneck is the review. Well, I have tried to point out this here [1].

Sometime ago, "maintainer" of submodule had been discussed. For
example, one could imagine that the commit access comes with the
"responsibility" to take care of a submodule; with great power comes
great responsibility. ;-)


[1] https://debbugs.gnu.org/cgi/bugreport.cgi?bug=38846#71


> I hope this makes sense. Maybe I am being overly nitpicky, I just really
> like clarity. :)

Me too, :-)


All the best,
simon