Storage internals: UUID

[Mark Grandi]

They might be talking about ghost revisions, where for some reason all
of the revisions are not present, but the history references them. Like
Daniel said the parents are obviously there when the revision gets
created so that isn't a problem, but if one were to verify the history
of a branch, and there are ghosts, then it fails, or does.something to
try and get the missing revs , etc. But despite that edge case I still
think that it would be good to add unless there are other problems
From: Daniel Carrera
Sent: 6/14/2012 5:02 PM
To: Aaron Bentley; bazaar at lists.canonical.com
Subject: Re: Storage internals: UUID


On Thursday, June 14, 2012 at 7:39 PM, Aaron Bentley
<aaron at aaronbentley.com> wrote:
>
> The problem with attempting to guarantee the history of a revision is
> that bzr doesn't have a static picture of the history of a revision.
> One repository might include all the parents of a revision.  Another
> might have only some of them.  And over time, new ancestral
> revisions may be added to a repository.

Oh...

Can you explain how it is that Bzr can have a revision without the
parents or why that is a good idea? I'm sure that this was a conscious
decision by the bzr developers, but I would like to understand it...
When you pull a revision but not its history, do you think it might be
possible to at least pull the SHA1 sum of the parent? I am assuming
that whenever a revision was made, the parents were present (it's hard
to imagine a successful commit otherwise). So I'm thinking that if you
pull a revision without its parents, you'd still download the
"history-aware testament"... I'd need to think a little about how the
details would work.


Cheers,
Daniel.