
Shae Matijs Erisson
Graham Klyne
writes:
You paint a fairly compelling picture. I think that the big advantage of darcs that you describe is that it can merge changes from repository to repository, rather than just from sandbox to repository. That seems useful to me, because I do use a locally-managed respository for all my work... it's the way I do backups.
Yep - currently, I must have my SVN repo locally on my laptop, and manually copy it to an NFS-mounted disk which is backed up regularly. With a distributed model, I could have repositories both places, and sync them when online.
"darcs inittree" creates a new repository in the current directory. "darcs get http://www.ScannedInAvian.org/repos/wikiwiki" sets up a local copy of a remote repository.
In case it isn't obvious, there isn't really any difference between a 'working copy' and a 'repository' in darcs (nor arch, IIUC). You 'get' the tree, and start working on it. Changes are 'record'ed locally, and 'push'ed (or 'pull'ed) from the other end) for synchronization.
- repository-hosting requirements?
filesystem, and an http server if you want others to be able to pull from your repository.
..over HTTP. You can pull if you have filesystem read access, and push if you have write access, too. (There's recently been some talk on the darcs list about some changes here, but I think this is still correct)
- ease of setting up secure connections to a remote repository?
You can use ssh[2] or sudo to push to a repository
Or use procmail and GPG-signed email. A bit complex, but there's good documentation with the necessary steps. -kzm -- If I haven't seen further, it is by standing in the footprints of giants