What is the managment overhead of decentrilized version control?

Adeodato Simó dato at net.com.org.es
Wed Jun 21 14:34:41 BST 2006

* John Arbash Meinel [Wed, 21 Jun 2006 08:16:59 -0500]:

> Simon Ekstrand wrote:
> > Unless this has been fixed recently, it's broken, atleast for
> > open-ssh(d). Openssh's serverside sftp implementation masks out any
> > suid/sgid bits on chmod operations for regular users. The sftp transport
> > mkdir method does a mode-less mkdir then a chmod, which doesn't work.
> > Including the directory mode with the mkdir call would work fine
> > (tested), but there's a comment in the mkdir method about this breaking
> > with the sftp server used for tests.

> > This problem bit me when setting up a central branch for shared
> > checkouts.

> If you already had the sgid bit, then wouldn't it force the creation to
> also have the sgid bit? Or is it the problem that the final chmod
> removes the bit that was already set?

> I have it working with standard opensshd, so I'm not sure why it isn't
> working for you. But it is definitely something that we want to support.

Isn't this all solved with ACLs? I need to set up a shared repository
soon, and was counting on ACLs saving my day. Worked fine in the tests I
did, with bzr push.

> The only times I've found it to fail was when I was using rsync to push
> my branches, which caused the remote end to take on the local
> permissions, rather than the shared remote permissions.

Right, I also suffered that. :)

Adeodato Simó                                     dato at net.com.org.es
Debian Developer                                  adeodato at debian.org
         Listening to: Fangoria - No sé qué me das [remix + de lo mismo]

More information about the bazaar mailing list