Storage internals: UUID

[Daniel Carrera]

On Thursday, June 14, 2012 at 5:31 AM, Mark Grandi <markgrandi at gmail.com> wrote:
>
> Sorry daniel, i keep forgetting to change the 'to' field to
> the mailing list rather then the person i'm replying to...
> 
> (real message)
> 
> I'm just really bad at explaining myself, but yes this isn't about gpg (totally
> at least) but it does seem to me that if we just put the parent's sha1 hash
> in the long form of the testament, we would be able to verify the history
> of the branch. Am I right in this assumption? are there any repercussions
> of adding this to the testament output?

I am not the best person to answer this, but from the responses I got on this list, this is what I gather:

1. They thought about using SHA1 sums + history as identifiers, as first done by Monotone. They chose not to. This decision allows Bzr to have a concept of "foreign branches". So, for example, you can treat an SVN repository as a foreign branch and all the SVN revisions get IDs without you having to do hashes for the entire history of the repository...


2. But you and I are not talking about using SHA1 sums + history as the main internal identifier... You and I are talking about the "testament" feature... I don't know, but at first sight I don't see any obvious problems with a "testament" that includes the SHA1 of the parents... You would have to store the testaments in some sort of index file, so it would require a small addition to the Bzr storage format, but that doesn't sound very difficult.


So maybe it is practical to lobby for a change to the Bzr format... Or more likely, the bzr developers might be interested in seeing a working plugin that implements the feature and if they like it they might turn it into a new feature.

I would be very interested in hearing from Bzr people here... Does anyone think that a "testament" that also guarantees the history of a revision is a *bad* idea?


Cheers,
Daniel.