bzr push preserve parent location?

[Martin Pool]

On 18 Oct 2006, John Arbash Meinel <john at arbash-meinel.com> wrote:
> Alexander Belchenko wrote:
> > Today I discover that bzr push is preserve parent location. Even if it
> > have no sense for pushed copy. Is this intended behavior?
> > 
> > -- 
> > Alexander
> 
> I'm not sure what you mean by preserving the parent location. Does the
> remote location already have a parent which it isn't changing? Is it
> setting the parent of the remote object to self? Is it setting the
> parent of the remote object to the parent of self?

Tangentially, I think we should probably get away from calling them
'parent' locations, because it can cause this kind of confusion.  It no
longer seems like a good idea to assume there's commonly a single place
you push to and pull from.  Instead 'remembered push location' or
'default push location'.

> As near as I can tell, we seem to set the parent of the remote object to
> the parent of self.
> 
> This only happens with the initial push, future pushes do not update the
> remote parent. (It happens because an initial push is actually closer to
> a 'clone' operation).
> 
> I think it makes sense to not set the remote parent at all. It might
> make sense to set it to self, but you don't know what route other would
> have to self.

Well, it would be consistent with 'bzr branch' if it set the remote
parent to point back to the origin.  Really it would clean things up
more if we fixed bzr branch to allow creation of a new remote branch.

When making a new remote branch we probably don't reliably know how to
set the parent in a way other people reading it will understand.
Perhaps we should instead add 'bzr branch --parent URL' and
'--no-parent'.

-- 
Martin