Repository size . . . is this right

[Talden]

>> If you run "bzr pack" twice without committing or pulling or adding new
>> revisions in any other way, it will exit without doing anything.
>
> I'd support having 'bzr pack && bzr pack' leave an empty obsolete_packs
> directory. I also completely agree that the current behaviour confuses
> people and thats *not good*. However, safety is really important too.

@Robert Collins
Though it doesn't sound like it's the intent of pack to reduce space
(it's about reorganising the packs for speed right?), there are
certainly cases where both are desireable.  Rather than an obscure
'pack twice' gesture to clear obsolete, how about a --shrink option on
the pack command that will also delete any unneeded content in the
.bzr after a successful pack (yes I know, if the pack makes a mistake
you're toast, so let's not make shrink the default).

>> > 300MB compared to 130MB is a bit of a disappointment.

@Russel Winder
If the obsoletes are included then they're 50% of the size after
removing all working-tree costs... Also, if git only has the one
working tree, do the same with bzr - use remove tree on branches
you're not actively working in.  It wouldn't surprise me to see that
bzr figure some down to 150-160MB.

> The split-inventory repository work includes hooking in better
> compression for inventory objects; During, or shortly after that, we
> should have better compression for texts too. Past experience suggests a
> 50% or better reduction.

@Robert Collins
I cannae wait... How is this work progressing?

--
Talden