Re: [PATCH 1/2] Maintainers.txt: update email address for Leif Lindholm

["Laszlo Ersek" <lersek@redhat.com>]

On 01/16/20 22:34, Leif Lindholm wrote:
> On Thu, Jan 16, 2020 at 19:54:43 +0100, Philippe Mathieu-Daudé wrote:
>>> Phil, would it make sense for us to ask Leif to post an update to
>>> ".mailmap"?
>>>
>>> For example, what happens if we run "git shortlog" over a period that
>>> contains patches authored by *both* of Leif's email addresses? Would
>>> those entries be merged into a single block? Would such a merging be
>>> desirable (that's for Leif to decide / propose)?
>>
>> One use of .mailmap is to clean/unify the NAME part.
>> In this case Leif isn't affected because his NAME is the same.
>>
>> The other use is to redirect email when a user switch email. When
>> fixing/referring old commit, some email tools might take the old commit
>> author/committer. In this case .mailmap helps. I think it is desirable in
>> Leif particular case.
> 
> The question is *why* we use the mailmap to begin with.
> It is not clear to me that there was an understanding at the point
> .mailmap was added that the intent was to provide a mapping of people
> to current email addresses in perpetuity.

It's one of the possible / nice-to-have goals, for "git shortlog".

It's absolutely *not* a requirement -- and I wanted to mention again
that such a cross-domain mapping is not something a maintainer can heap
on top at a push. Cross-domain generally implies cross-employer, and
companies might easily not be OK with that. So it's fully up to the
person to propose (post) such a mapping. If they don't, then there's not
going to be such a mapping; end of story :)

> Indeed, the initial commit message says:
> "The .mailmap git feature helps fixing commit mistakes (in name/email)."
> 
> The comment header in the file itself says:
> # This list is used by git-shortlog to fix a few botched name translations
> # in the git archive, either because the author's full name was messed up
> # and/or not always written the same way, making contributions from the
> # same person appearing not to be so or badly displayed.
> 
> Now, I'm not saying that providing a way to track individuals across
> employments, acquisitions, or other transitions is a bad idea. And
> using the .mailmap is a workable way of doing this, since it needs to
> be manually invoked outside of shortlog (it could be disastrous if it
> didn't, or the default git behaviour changes in the future).

I agree: ".mailmap" must not unsolicitedly collapse email addresses.

> *But* before we start treating it in this way, I would like to see the
> intended use of .mailmap documented (in the tree), and that change
> reviewed by at least a couple of stewards and enough others to be able
> to say we have a consensus.
> 
> If we do, I have no objections to the below patch being added at that
> point - but from my viewpoint, doing it before then would mean an
> unreviewed change to the process as some people understood when the
> file was first added.

I don't think we're actively proposing cross-domain mappings, as part of
the process. In my mind it's something everyone is "free to consider". I
just thought you might want to consider it.

I'm also not against formalizing this option, but I don't feel like
actively pursuing it.

Thanks!
Laszlo