[RFC] proposed user doc for nested trees

[Ian Clatworthy]

Aaron Bentley wrote:
> Aaron Bentley wrote:
>> Since the code isn't updating any extra info, updating branch reference
>> locations would be a big burden, and you'd need to ensure that it was
>> impossible to move a checkout any other way, to avoid breaking the
>> user's tree.  That's a very difficult guarantee to provide.
> 
>> Whereas, moving checkouts around is a fairly rare event.
> 
> What I mean is that moving subtrees within a working tree would be much
> more common than moving an entire set of nested trees (composite tree)
> around.

I disagree. I regularly do something like:

  bzr branch trunk foo
  (hmmm - bar is a better name for that branch)
  mv foo bar

That needs to work if foo has nested trees. *I* think we want
lightweight checkouts to use relative paths just like Alexander
suggests, for this case and many others.

> I think that means we should optimize for the first one, but it doesn't
> mean we can't support the second.  For example, perhaps "bzr switch" in
> the root tree could update the branch references.

Using "bzr switch" like this is a workaround. It isn't "Just Works" IMHO.

Ian C.