Re: rewriting history; Was: Concerns/questions around Software Heritage Archive

[Andreas Enge]

Am Mon, Mar 18, 2024 at 04:33:49PM +0200 schrieb MSavoritias:
> Actually gitlab already is facing something like that and they are doing
> what was proposed elsewhere: mapping of UUIDs to display names
> https://gitlab.com/gitlab-org/gitlab/-/issues/20960

Interesting, thanks! It is something that maybe could be implemented by
Savannah, but it would probably require a bit of thought. And yet again,
somehow the mapping uuid<->"real" names would have to be public (people
would "git clone" commits with uuids, and would need to locally replace
them by "real" names); so people can always keep copies of the mapping
over time.

I am also not quite sure about the signing process for committers;
in principle keys are enough, but in GPG they are tied to email addresses,
and I do not know whether we use this in Guix.

In the end, my impression is this will not achieve much more than what we
already have with the .mailmap approach. In a sense, everyone would use
a pseudonym (their uuid), and then we would keep a mapping between these
pseudonyms and, well, "real" names or other pseudonyms chosen by the
contributors...

Hm, this could indeed be implemented exactly with .mailmap, no?
We would need to enforce that authors use a uuid of a specific format,
and potentially an empty or dummy email address, or another uuid.
Then we could keep a .mailmap file. The history of "real" identities
would still be visible in the git history, but as said above, anyway
we could not prevent people from storing the association information
over time.

> Right fair. As I have said before SWH does break Guix CoC effectively right
> now.
> So what Guix does from this point on will effectively dictate if the CoC is
> valid or not.

Well, the CoC is valid on our communication channels; so what SWH does with
our software is outside its scope (that is governed by the license).

Andreas