repositories and working trees...

[Erik Bågfors]

Hi

A discussion came up on IRC today regarding repos, the current
init-repo command creates repos without working trees.  Of course I
wanted a repo with branches with working trees...

I think there are two use cases here.  One is the normal shared repo
where one or many developers have a number of branches.  This can be
on a server for example.  This is the way most people think of
repositories.  In this case no working tree is expected.

The other is when I have 10 bzr branches that I want to keep on my
computer. In this case, it's very usefull to have repos with branches
with working tree...

For example

~/src/bzr is a directory there I collect all my branches
~/src/bzr/bzr.dev is the bzr.dev branch
~/src/bzr/bzr.erik is my private branch
~/src/bzr/bzr.somethingelse is another branch etc.

If ~/src/bzr is a repo, it means that I can collect my branches as
usual, work on them, like they were stand alone branches, etc.  but
that requires the branches to have trees.

The question came up on what bzr should support, and what should be
default. Here is my take on this

1) bzr init-repo should by default do what it does today.  This will
support the first use case, which is what most people will expect from
repositories
2) bzr init-repo should take a flag to allow doing working trees.


Another important thing, that I haven't tried but that I expect works,
is that if I have a shared repo that allows working trees.  I should
be able to do "bzr checkout sftp://xx/xx/xx" and the repo will be used
for the bound branch that's created.  This is how I expect to be
working alot.

Also.. (somewhat off topic here)
There should be some way of listing branches in a repo. For example
"bzr list-branch sftp://xx/xx/xx" should list all branches there.


/Erik