[RFC] quickbranch

Martin Pool mbp at canonical.com
Thu May 25 05:18:36 BST 2006


On 24 May 2006, Aaron Bentley <aaron.bentley at utoronto.ca> wrote:
> Martin Pool wrote:
> | On 24 May 2006, Aaron Bentley <aaron.bentley at utoronto.ca> wrote:
> |
> | Well, like changesets, but possibly optimized more for compactness and
> | fast random access to their contents, rather than human readability.
> | Perhaps this is just a different format of changeset?  Or perhaps using
> | them just as they now are would be a good first step.
> 
> I'm not sure why random access is desirable, but changesets definitely
> aren't random-access-- almost all revisions depend on other revisions in
> the changeset.  They're also slow when you have large numbers of
> revisions.  (The scaling looks fine, they're just a bit slow.)  One
> possible enhancement would be to use the patches directly to generate
> the knits.

I was getting ahead of myself: for "reference storage" of history you want
to fairly quickly be able to recreate a particular version of a
particular file.  If you used changesets but then integrated them into
e.g. knits that would be fine.

-- 
Martin




More information about the bazaar mailing list