[RFC] bzr.jrydberg.versionedfile

[Martin Pool]

On 21 Dec 2005, John Arbash Meinel <john at arbash-meinel.com> wrote:

(happy holidays! :-)

> > Well, even with the above constraint that it is only about the actual
> > branch, I don't know if you gain much. If the branch keeps track of "I
> > am this revision", then any complete revision graph can give you the
> > full ancestry for that revision.
> > And if we change so that 'revision-history' is always the first parent
> > along the history (so converging two branches actually means that one of
> > them 'wins'), then a branch is defined purely by a single revision-id.
> > All of its history is encapsulated in that single identifier. (tags may
> > not be, though).
> > 
> > In general, I think revision-graph is a non-starter. Simply because the
> > information is already available elsewhere.

It seems like it's an optimization for format which don't let you
quickly retrieve the graph from the actually stored revisions.

> I have to modify this statement for 1 possible use case. If you wanted
> to have a limited history branch.
> 
> Specifically, if you wanted to create a branch which *thinks* that it is
> a full branch, but doesn't actually contain some of the ancestry. At
> present, I don't think there are any plans to support this. 

So, in particular, having the revision-history would let you know what
revno you're on, even if you don't have all the mainline revisions up to
that point.  That might well be useful, since the UI tends to assume  we
do know the revno.

As for the idea of the revision-graph, it's probably good to have
revision-history as a cache if it's not really fast to get it from the
canonical storage.

> Considering how much Martin has fought about not wanting Ghosts,
> though he accepted because he wanted easy upgrades from Arch.

Well, I think in saying we can refer to not-present or
not-really-existing revisions *at all* we have taken most of the pain.
The extension here would seem to be mostly in allowing things on your
main-line history to be absent as well.

But perhaps cheap checkouts are a better approach for this kind of
thing.

-- 
Martin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060110/76bb87dc/attachment.pgp