Re: [Gnu-arch-users] Why we might use subversion instead of arch.

[Miles Bader]

On Fri, Feb 20, 2004 at 04:45:54PM -0700, Pierce T. Wetter III wrote:
>  The main rough edge I see is that I think local-mirrors of a remote
> archive should allow you to commit, with the changes propagating upwards to
> the source. Especially a greedy mirror.

I agree, `write-through' mirrors would be a fine feature; I'd probably use
mirroring more if it were available.

I suppose the simplest version would be simply keeping two archive paths for
each archive name, saying `use path A for commits, and path B for everything
else'.

However, I think it would be desirable to maintain a stronger illusion of
consistency between the two archives (the real archive and the mirror), and
I'm not sure what problems one would run into trying to do that.  One could
do what you suggest, and make a `greedy' mirror -- updating the mirror at
every opportunity -- but of course even checking to see if an update is
needed takes _some_ time, and one nice property of truly local mirrors is
that they're always really fast.

Additionally, this model (commit remotely, then mirror back the changes to
your local mirror) entails transferring committed changesets twice, so I
wonder if there's any simple/sane way to commit to the local mirror locally
at the same time as committing to the remove archive (lock remote, lock local
mirror, commit remote, commit locally?)

Well, now I'm just babbling...

-Miles
-- 
I'd rather be consing.