Re: updating async on elpa

[Stefan Monnier]

> Having one repository per package is how it works everywhere else
> (Savannah, etc.), and is easier to understand because of that. This is
> from the point of view of a contributor of course; I cannot speak to how
> it looks from Stefan's side.

>From my side it's plain and simply not an option at all because creating
a new repository on git.sv.gnu.org requires manual intervention from the
Savannah admins.  Having had to design a solution that lives within this
constraint, I'm actually pretty happy about the current result, so if
the constraint were lifted, I'd probably want to try and preserve
the advantages.

In any case the `:auto-sync` feature should make it much less of an issue.

> I would be much happier if emms and rt-liberation had their own
> repositories since it would mean that I couldn't possibly break anything
> except emms or rt-liberation when I push to those projects in elpa.

The `elpa.git` is configured to reject non-fast-forward pushes, which
should catch most blunders.  Plus it is cloned at various different
places which are not all constantly updated, plus (almost) all changes
to it are reflected in the elpa-diffs mailing-list.  So even if you do
manage to make some really terrible mistake it should be quite easy to
recover from it.

You might want to activate the `:auto-sync` on your packages, tho to
save you from the worry.


        Stefan