Are read-only stores a bug?
aaron.bentley at utoronto.ca
Sun Jun 19 20:23:24 BST 2005
-----BEGIN PGP SIGNED MESSAGE-----
The bzr design is supposed to make it really cheap to branch. You can
use cp -a to do it, and start committing right away. Unlike Arch,
branches don't require you to pick a name and assign that name to the
branch forever and ever. These are good attributes.
The net result is that you wind up with a lot of branches, many of which
you no longer need. I find I'm often deleting branches because their
changes have been integrated, or because they didn't work out. And some
are test data that's no longer useful.
But I can't use rm -R to delete them. Not sanely anyway, because there
are two readonly files for each revision, and one readonly file for
every set of contents a file's ever had. So rm -Rf it is.
But rm -Rf is a really bad thing to commit to muscle memory. Only a few
commands come to mind as more dangerous (dd for example).
The likelihood of anyone venturing into .bzr and accidentally deleting
the store contents seems very remote. So I don't think the stores
actually need to be read-only, and this would avoid teaching users to
type rm -Rf routinely.
If we wanted to keep the stores readonly, I suppose we could have a "bzr
unbranch" command that would only delete bzr branches.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
-----END PGP SIGNATURE-----
More information about the bazaar