On Tue, Oct 30, 2007 at 08:06:18AM -0400, Aaron Bentley wrote: > Jelmer Vernooij wrote: > > On Tue, Oct 30, 2007 at 07:43:19AM -0400, Aaron Bentley wrote: > >>> would it perhaps be possible to have a data format that can > >>> store rich roots but not subtrees and as such doesn't have to be > >>> marked experimental? > >> But it would still be a watershed, > > Why does it have to be a watershed? > With non-rich roots, there's no way of storing the last-modifying > revision, so when necessary, we pretend that it changes for every > revision. With rich roots, it changes extremely rarely. There's no way > of representing that in a non-rich-root format. Ahh, thanks, I didn't think of that. > > Would it be possible > > to simply limit downgrades to revisions that have their tree root set > > to bzrlib.inventory.ROOT_ID. > The only revisions that could be downgraded would be those that were > originally converted from xml5. Hmm, so that basically makes that idea of allowing limited downgrades impractical. I would still like to propose a --knitpacks-richroot format or something, for use by bzr-svn, if subtrees are some time off. The watershed has to happen at some point in time anyway, may as well be now. Cheers, Jelmer -- Jelmer Vernooij <jelmer@xxxxxxxxx> - http://jelmer.vernstok.nl/
This is the launchpad-users mailing list archive — see also the general help for Launchpad.net mailing lists.
(Formatted by MHonArc.)