Re: Improving GNU ELPA

[Stefan Monnier]

>> - we don't want to fetch from non-GNU servers, so we need the maintainer
>> to push to elpa.git explicitly.
>
> Not really (I think you hint at this a bit later): if the FSF hosts a gitlab
> instance, or any sort of (multi-repo) git hosting, then developers could
> just mirror their repos to that instance.

Indeed, by "elpa.git" I really meant "some repository under our
control".  Currently it's elpa.git, but that could be expanded.

> And in fact we already have this.  ELPA could work just like MELPA, but
> restricting the package source to Savannah.  Then publishing to ELPA would
> be just like MELPA, except for having to mirror to Savannah from time
> to time.

Note that allowing any package on Savannah would already be quite
different, since people without copyright papers have write access to it
(and we'd lose the write access for Emacs maintainers, as well as the
elpa-diffs reviews, ...).

I guess it would be marginally better than allowing any package from
"anywhere" (e.g. when a package goes unmaintained, there's a process
that could allow us to get write access), but it would add the hurdle of
being accepted into Savannah, so I don't think it would eliminate enough
friction to make a significant difference.


        Stefan


PS: BTW, to clarify my position: if it were up to me, I'd get rid of the
    copyright assignment policy for GNU ELPA (and for Emacs as well,
    while we're at it), but I'd keep the "locally hosted in a repository
    to which we have write access, with commit-diffs".