bzr-git, the git map, fetching revisions, and being very slow

[Martin Pool]

2009/11/23 Jelmer Vernooij <jelmer at samba.org>:
> On Sun, 2009-11-22 at 09:19 +0900, Stephen J. Turnbull wrote:
>> Jelmer Vernooij writes:
>>
>>  > It's unfair to compare imports from git to bazaar with native clones
>>  > between git repositories. Compare with pulling from a git repository to
>>  > a git repository using bzr instead.
>>
>> Of course it's not unfair, because the comparison we should care about
>> is performance in the user's eyes. The user may be willing to make
>> the tradeoff, but the more we say about why they should in the face of
>> obvious defects less they'll trust us.  Note that that's different
>> from *explaining* the performance differential, such as "This
>> operation is sufficiently rare that we put low priority on optimizing
>> it, and it's a relatively difficult task since we are Bazaar experts,
>> not git experts.  You should not extrapolate and assume all bzr-git
>> interactions will have such low performance."
> My point was that copying data from one repository in a particular
> format into another repository in the same format is *always* going to
> be faster than copying into a different format because there would be
> some sort of conversion involved in the second case. It is never going
> to be as fast.
>
> Perhaps we should be telling the user that an operation is expected to
> be slower when fetching between differing formats.

I think we should, or at least something like "translating from
repository format 'bzr-git' to '2a'".  This also comes up when going
from 2a to pack formats or vice versa: we recompress all the data so
it's unavoidably going to take a while, but just giving a clue as to
why it's slow would help.  I think there is a bug for this.

-- 
Martin <http://launchpad.net/~mbp/>