[Gllug] subversion

[Jan Kokoska]

On Tue, 2004-12-14 at 08:18 +0000, Richard Jones wrote:
> On Mon, Dec 13, 2004 at 11:40:27PM +0000, Jan Kokoska wrote:
> > Jan / one happy CVS user who might consider Subversion, next month,
> > maybe
> 
> The real problem with Subversion (and CVS) is that they define a
> priviledged class of "committers" who are allowed to use the
> repository in read/write mode.  Everyone else gets to checkout and
> update from the repository at best but basically are excluded from
> using version control features at all.
> 
> It turns out that the priviledged class ("committers") isn't a
> necessary requirement at all.  If everyone has their own repository,
> which is basically how darcs/arch/bk work, then everyone can have the
> version control features.  Changesets can pass between repositories to
> synchronise them in a controlled manner.  It's then just a matter of
> agreement which repository is "the" repository, or primary source of
> the code.
> 
> The practical upshot of this is that I can develop free software and
> it's much simpler for others to join in.  If I find another developer
> willing to help out, I don't need to mess around with ssh keys giving
> them a "committer" account on my CVS server.  Consider that most
> developers will stick around very briefly, some will pop up from time
> to time, and only a very few will become dedicated to providing
> patches to the project.  I now no longer need to worry about which of
> these will become "committers", but instead concentrate on whether or
> not to accept individual patches.
> 

I perfectly understand the argument. It's just that all my development
is commercial at the moment and there is no trouble defining this class
of "commiters" as our employees are pretty stable (and when the
by-products are open sourced, which happens regularly, it turns out
hardly anybody even checks it out of CVS, not to mention e-mail about
contributions.. usually they don't work at code level at all).

A number of people is using versioning systems simply to archive *their*
code (and possibly even config files and homes), as a good practise.

CVS is total crap in merging.. well turns out even that I need very
rarely.

So if you are talking about a project size of kernel, fine, but let me
say I am happy not working on that -- mind bogglingly massive.

Are you working on any project commercial or libre, that does actually
need darcs/arch/bk?

Jan

-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug