Comment 1 for bug 88848

LockableFiles itself may be an unnecessary abstraction, but at minimum, we ought to keep the lock itself abstract.

I think locking mechanisms should be the purview of the control directory format. So the hpss locking can be different from the http locking, and if we need to create a new lock mechanism at some point, we only need a new BzrMetaFormat-- we don't need new tree, branch and repository formats.

So I think the BzrDirFormat should control the locking mechanism used.