On Fri, 2007-02-09 at 14:18 +0100, Magnus Holmgren wrote:
> Well, Subversion tries to be like CVS in terms of basic usage (checkout,
> update, commit), but I think the fact that changes to the source tree as a
> whole are versioned, as opposed to single files, is at least an important
> improvement, although not as revolutionary as the distributed model. Moving
> to Subversion should be possible in a very short time, so it's not much of an
> upheaval.
Moving to git would also be possible in a very short time. It's trivial
to import repositories from CVS to git...
http://git.infradead.org/?p=users/dwmw2/exim-src;a=summary
http://git.infradead.org/?p=users/dwmw2/exim-doc;a=summary
I've never been bothered by atomicity of commits per se; I've used CVS
for a lot of years and it just hasn't been a problem for me. But the
_grouping_ of individual changes to different files into logical commits
is nice, I agree. You can retrieve this information from CVS with
'cvsps', and in fact when you import a CVS tree into git it does just
that for you. See, for example,
http://git.infradead.org/?p=users/dwmw2/exim-src;a=commit;h=32c23385a5ee71529375ddf26d8acfbeaecbda70
I find the distributed model is very useful. It allows anyone to put up
their own 'branches' in which they can develop, share and debug a new
feature, which can then be easily merged back into the main tree.
(Those git trees are just an example and won't be automatically updated
from cron like some of the other CVS->git shadows I run; I'd need to set
up some method of read-only CVS access for that.)
--
dwmw2
