Bazaar dirstate locking
Stephen J. Turnbull
stephen at xemacs.org
Sun May 16 01:35:48 BST 2010
Ben Finney writes:
> "Stephen J. Turnbull" <stephen at xemacs.org> writes:
>
> > Bazaar is a *distributed* VCS. It is a database designed for
> > concurrent operation, with database operations conducted locally on
> > ACID, then propagated as completed transactions to other nodes of the
> > database. If you insist on operating on incomplete transactions across
> > the net, you're going to pay for the privilege.
>
> A wrinkle in this analogy is bound branches. As I understand it, a
> modification on a bound branch causes the VCS data to be written to the
> *remote* repository first, and only on success is the same write done
> locally.
>
> I don't know how that fits with the model you espouse above, or whether
> you consider the local and remote branches to be separate nodes.
It doesn't fit with that model. Fixing the model is hard, not as hard
as fixing the speed-of-light bug in interstellar travel, but hard
none-the-less.
The exception is that especially privileged developers could be
allowed to have bound branches on the same LAN as the upstream branch.
More information about the bazaar
mailing list