Branch?

[Paul Moore]

On 9/7/06, Russell Suter <maverick at simlogix.com> wrote:
> Giorgos Keramidas wrote:
> > The easiest of the two, by far, is to simply "clone" the original
> > repository and start working in the cloned copy immediately.
> > Since the clone of the original repository is a standalone,
> > independent repository of its own, whatever changes are made
> > locally to the clone and committed as new changesets do not have
> > any immediate effect on the original repository.
> >
> > Only if you explicitly do a 'push' command, the local changes
> > will get a chance of being applied to the original repository.
> >
> > This is why the shoothing mandra says that "a clone is a branch" :)
[...]
> Ah!  Lights on now.  I've overcome yet another CVS-ism.  I was thinking
> in terms of a branch tag but I don't even need a tag to branch.  So when I
> clone, the tip of my repository is automatically a branch point.

A comment from another newbie...

One thing that still prevents me from being completely soothed by the
mantra is that (AFAICT) cloning doesn't have any metadata associated
with it. So, for example, I can't branch (clone) and mark that point
in the history with a log message "Branched to try out XXX".

Instinctively, clone-as-branch feels to me like my pre-VCS approach of
copying the working directory to try something out (a process which I
usually managed to mess up in one way or another). In that context,
clone-as-branch leaves me feeling a bit too exposed to my own
incompetence for my own liking. [I know it's not *really* as bad as
just using file copies for versioning, but logic never overrides gut
feeling :-( ]

What I'd like is a way of working which allowed me to use clones for
branches as at present, but be able to "check in" these clones to some
central repository, so that one place at least had a record of
everything I'd done, including all branches, so I could back that one
up and feel completey safe (barring offline working, of course).

Paul.