anyone using subversion?
Roger Oberholtzer
roger
Sun Jan 2 07:04:48 PST 2005
On Thu, 2004-12-30 at 15:35 -0800, Net Llama! wrote:
> On 12/30/2004 02:24 PM, James McDonald wrote:
> > Net Llama! wrote:
> >
> >> Anyone out there using subversion (instead of cvs)? If so, i've got a
> >> few questions.
> >>
> > I started using subversion a couple of months ago to keep documentation
> > and websites versioned. Use the apache server web access method to
> > access the repositorie and the windows client from work to send receive
> > versions. Don't use it in "anger" but have found it to be easy to use.
>
> Does subversion have the same limitation as cvs where once you 'remove'
> a file from a repository, it gets thrown into an Attic subdirectory, and
> cannot be re-added with the same filename again unless the Attic version
> is manually deleted on the server first?
I do not think so. One thing we like about subversion is that it does
indeed track when something is removed. However, you need to explicitly
remove it. That just means that when it dumps the tree, that thing is
not included. When you add it back, it is there again. I think cvs and
the like were odd about this in that it was an added hack after the
design was complete. sccs/cvs/rcs never had the concept of something
leaving.
Subversion is a big step forward from cvs-type systems. The only thing I
always feel uncomfortable about is that each tree is kept in one single
database. So, if you have a 6 gb source/tools tree (as we so), that
makes for a gigantic database. Luckily, our source is packaged in
smaller units, which lend themselves to separate subversion databases.
The other thing we like about subversion is that there is no locking.
You check out something and it is yours. Checkouts are unlimited. Anyone
(who has access) can check out as many times as they like. The control
is done at checkin. This is a brilliant step forward from cvs
checkout-time locking. This does great things for multi-developer
multi-location development. Something cvs never really did. Unless you
used dozens of shell scripts to layer some sort of control system on
top. Not the real way to go.
--
Roger Oberholtzer <roger at opq.se>
More information about the Linux-users
mailing list