On Thu, Aug 19, 2021 at 6:29 AM Ivan Kelly <iv...@apache.org> wrote:
>
> > My two cents as a mere contributor who once wrote a PIP. I would love to
> > see PIPs as a versioned document in git instead of the Wiki. This would
> > provide history and context, make it easy to comment and propose fixes and
> > enhancements which is currently not possible with the Wiki. The PMC would
> > still have the possibility to accept or reject PIPs by merging or not to
> > master with the usual and well-known PR review process.
>
> +1 on this. Wikis, despite being collaborative, often end up being
> write-once-and-forget. If the PIP is a PR into the repo, it'll hang
> around like a bad smell until accepted or rejected.
>
> -Ivan

Ivan / Cristophe,

I'm actually proposing to use GitHub issues instead of Wiki from now on.

In my opinion, the problems with using Pull Requests for this scope are:
 * It blurs the line of what the approval and merging of the PR and
the approval of the proposal, which I believe should still be done on
the ML.
 * We shouldn't discard proposals that are rejected, instead it would
be very good to keep them, and the context of the discussion that
happened, for reference. (eg. a new proposal might come up with a
different way to solve a problem that was already discussed).

GitHub issues are not perfect, for example they lack the versioning of
the content, but have some good parts:
 * Anyone can create issues (not only committers like it's the case for wiki)
 * We can attach labels to issues
 * Users can comment on the issues, so even people that are not on the
dev@ list will be able to add their feedback
 * We can still keep an index of all the proposals in the Wiki
(curated by committers group)

It was also raised by Lari that if we allow edits to the issue
description we need a way to ensure the vote is done on a specific
version.
For that, we could ask to paste the "final" markdown in the vote email thread.

There might still be minor adjustments allowed after the vote (eg: to
reflect correctly the reality of the implementation vs the initial
design).


Matteo

--
Matteo Merli
<mme...@apache.org>

Reply via email to