New VCS Choice; SCM SIG

[Ralf Corsepius]

On Mon, 2007-01-22 at 22:27 -0800, Toshio Kuratomi wrote:
> On Tue, 2007-01-23 at 06:59 +0100, Ralf Corsepius wrote:
> > On Mon, 2007-01-22 at 21:46 -0800, Toshio Kuratomi wrote:
> > > On Mon, 2007-01-22 at 21:07 -0600, Josh Boyer wrote:
> > > > On Mon, Jan 22, 2007 at 06:33:17PM -0800, Toshio Kuratomi wrote:
> > > >  
> > > > > Right but we're talking at cross-purposes.  What I mean is, are we using
> > > > > these now with the fedora-extras and fedora-core dist-cvs?  If so, what
> > > > > are the commands you're running to do that?
> > > > > 
> > > > > Although we talk about branches with dist-cvs, there's no cvs branches
> > > > > in the repository that I'm aware of.  So I'm wondering if Karel has some
> > > > > commands that he's using now or if this is a "If we use real branches in
> > > > > the next SCM, it'll give us the ability to do operation Xyz in a better
> > > > > way."
> > > > 
> > > > Well, there's nothing to really prevent someone from using a branch for
> > > > private development.  The buildsys won't build off of it, but you can
> > > > commit stuff to a branch if you'd like.  And of course, there's also the
> > > > tagging we _do_ use.  Which is similar to a CVS branch.  You can actually
> > > > do the merge command using tag names too.
> > > > 
> > > > I can see some use cases for using branching with git, simply because
> > > > branches are trivial to create and work with.
> > > > 
> > > Yeah -- branches in subversion and bazaar are trivial,
> > How comes you consider branches in CVS to be more complex?
> > 
> > I for one (Many year's of CVS power-usage), have never found subversion
> > branches easy to use (I've never used bazaar nor git).
> > 
> cvs doesn't separate the concept of tags and branches.
Right, this bloats the repos sizes on the server (svn pushes this bloat
to the clients) and renders "branch removal" a pain on the server.

>   It also works on
> individual files rather than whole trees.
Right, but I don't see this as a disadvantage. It's a different working
principle.

>   It could just be a limitation
> of my brain but those two things made branches a lot harder for me to
> understand in cvs.
Well, on the client-side, branching is easy:
cvs tag -b branch-tag
cvs up -r branch-tag -dP

Unlike with svn, diffing between branches and getting to know about
branches (cvn log) is trivial.


What I find arguable/questionable in FE's CVS-repos, is the way
"FE-branches" have been implemented into it. They are implemented as
separate directories instead of CVS-branches. If real CVS-branches had
been used several details would have been much easier.

In the early days of FE, I had been told the reason for this design
decision had been AVCs, because CVS storing branches in files would
prevent AVCs to be applicable.

Ralf