[MERGE/RFC] Userdoc Driven Design on the Bazaar 2.0 UI

Matt Doran matt.doran at papercut.com
Fri Apr 17 02:13:49 BST 2009


Mark Shuttleworth wrote:
> Matt Doran wrote:
>> If you make implicit links between branches that are not completely 
>> obvious there will be pain!
>
> Agreed. The branch process should NOT be implicit about this. And it 
> should be possible to query a repository to get an idea of which 
> branches are using it.
>
> Mark
Has anybody considered the option of using a shared repository 
automatically created in ~/.bzr/repository/.  It would be created by a 
branch or pull operations, etc.  The default would be to always use 
this, unless explicity asked for via a "bzr branch --standalone" option 
or something.  All branches would be stored in this repo, local 
branching would always be super fast, and you'd get efficient disk 
space.  There is a downside that the repo is decoupled from the branch 
... but the behaviour would be consistent and explainable.  Also one 
global shared repository might be a bit unwieldy (e.g. formats, garbage 
collection, etc).

Another variation would be to store multiple shared repos in the home 
dir. e.g.

    ~/.bzr/repositories/<uuid1>/
    ~/.bzr/repositories/<uuid2>/

Where there was one shared repository for each "project".  I.e. one repo 
would be used for all branches of the "bzr" project.... another might be 
for branches of "emacs".   Do branches have some sort of identifier that 
uniquely identifies them as being a descendent of some "root" branch?   
Subversion has a per repository UUID .... does bzr have something similar?

It's a tough problem to solve elegantly .... I just thought I'd throw 
this one out there.

Cheers,
Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ubuntu.com/archives/bazaar/attachments/20090417/b7b86415/attachment.htm 


More information about the bazaar mailing list