On Tue, Apr 22, 2008 at 1:32 PM, Johannes Waldmann < [EMAIL PROTECTED]> wrote:
> What's the current best practice for setting up > a repository for remote write acces > (and not giving full shell access) > > It seems ssh is required, so scponly does not work. > This is quite the show-stopper. > > For CVS, one could apparently use (in sshd_config) > Match User foo ; ForceCommand cvs server . > Does something similar work for darcs? I'm afraid not: > > darcs' RemoteApply contains: > pipeDoc_SSH_IgnoreError addr ["darcs apply --all --repodir > '"++path++"'"] bundle > > while the corresponding function in cvs (client.c:start_rsh_server) > constructs the command "cvs server" > (literally those two words, no options, no arguments) > and this makes ForceCommand work. > > Perhaps this method can be considered for darcs as well. > The data will be piped anyway, so it should be enough > to prepend this with one line giving the path information. One thing that could work, although may not be optimal, is that you could use ftp or rsync to keep the repositories on the server side and then always do your work locally. So you interact with other people's server side branches, work locally and when instead of 'darcs push' or 'darcs send', you execute 'darcs-sync.sh' that runs your favorite (or allowed) copy program to clone your local directory on the server side. I'm curious to see other solutions people have for this case as mine is a bit silly. Jason
_______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
