First create the directory to hold the repository, eg,

 cd public
 mkdir cvsroot


Next give John and Peter write and read access.  Write access is
needed even just to checkout the files.

 fs setacl -dir cvsroot -acl jchill rlidwka
 fs setacl -dir cvsroot -acl pwp rlidwka


To use cvs locally, set the repository environment variable 
(eg, interactively or in .cshrc)

 setenv CVSROOT /afs/cern.ch/user/g/gmoor/public/cvsroot

(alternatively, specify the repository with the -d option to cvs
at appropriate command lines.  Because the repository is local we
don't need to specify an access method)


Now create the base repository

 cvs init


Now, we have already created a local working copy of the source
file tree from which we want to import initially the first generation
of files into cvs.  CVS works by assuming the current working directory
is the one which needs to be synchronise with the repository for any
given command, so we now need to cd to the local working copy, and
then issue the import command.

 cd /afs/cern.ch/user/g/gmoor/public/sctdaq

 cvs import -m "Original sources imported, post Aug99 testbeam" evaluation/sctdaq sct start


Note I have created an evaluation sub-repository;  when all co-authors are
happy with a version, we should also import into a production version.

This evaluation copy can now be checked out.  Note:  to use CVS we must
always poistion ourselves in an appropriate workspace before issuing
cvs commands.   In this case,  we should move to a directory other
than the one from which we imported the sources, checkout the sources
from the repository, then do a diff on the files to check all was well.
We can then delete (or release) both working copies.  Eg,

 cd ..
 mv sctdaq sctdaq.original
 cvs checkout evaluation/sctdaq
   (creates a new working directory evaluation/sctdaq)

 diff -r sctdaq.original evaluation/sctdaq

 rm -rf sctdaq.original


Note:  you can checkout the files locally as above on a machine which gives
you both read and write access to the files on AFS, or remotely using
the client/server mode of CVS.  This is indicated by putting an access method
other than :local: in front of the CVSROOT variable.  By default, with method
:server:, CVS tries to use an inbuilt version of rsh.  Rsh is not permitted
to run on atlas.wgs (and presumably other centrally support machines at 
CERN) so one must use ssh.  This is done by telling CVS to use a substitute
remote shell, by using method :ext: and by setting the CVS_RSH variable:

  setenv CVSROOT gmoor@atlas.cern.ch:/afs/cern.ch/user/g/gmoor/public/cvsroot
  setenv CVS_RSH ssh

where ssh is in my standard path on my local host.

With these environment variables I can checkout, update and
commit from Melbourne Unix server to atlas.




To use CVS from Windows or NT, you need the appropriate versions
of CVS and the secure shell, SSH.


The SSH implementation must be a transparent one, not an interactive
one.  I have found that the one downloadable from:

www.bmrc.berkeley.edu/people/chaffee/winnutil.html

does the job.