nested tree remaining questions
Aaron Bentley
aaron.bentley at utoronto.ca
Wed Feb 8 22:23:49 GMT 2006
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Robert Collins wrote:
> Seems to me there are three options :
> * an unversioned 'thing' on disk, present in [say] any or all of repo,
> branch, tree, and consulted in reverse order.
> * a versioned file - thus essentially present in the tree,
> * versioned data - like the .bzrignore plan a tree version of the
> 'current' data would exist too.
I think you're missing an option: data that is verisioned, but not in
sync with tree data. An example of this kind of data is tags.
I think it is useful to be able to update the child locations, as other
people's branches/repositories move around. This update should apply to
all historical versions, also.
It appears that child locations are not necessary for repository
branches. We could simply require that all nested branch revisions are
stored in the same repository. And with revision-history going away, a
branch will be just a revision-id.
> So the questions to disambiguate are:
> * Are historical versions of the data relevant? Yes - if I had a nested
> branch a year ago but do not now, even if I clean up the hints, a
> 'branch -r -1000' operation should be able to access them. This rules
> out 'unversioned thing'
Yes, but:
* will I want to update the data that applies to old revisions'? Yes -
If I nest PyBaz and then ddaa changes the location of his main branch, I
want people to still be able to check out any revision.
But also:
* Do I want to update my local settings without others being affected?
Probably. If I set up a local mirror, of ddaa, I may want to point at
that, instead.
Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFD6m900F+nu1YWqI0RAi4TAJoCitU/RQguP5YaPIFP83TxsRElIgCfUk72
6/n/CNwagyO73p9t0nFciSU=
=JmC0
-----END PGP SIGNATURE-----
More information about the bazaar
mailing list