Am I crazy for thinking about forwarding the DBUS session over an ssh connection much like the X session is forwarded? I've proven the concept with socat (to use the unix socket that dbus normally listens on rather than reconfiguring dbus to use a tcp socket) and it works just as expected. Any reason we wouldn't want openssh doing this natively? b.
Any reason why this has to be an unique feature? Is there some form of authentication cookie or something that needs to be managed? That is really the only reason why X forwarding exists. Otherwise it could also be yet another port tunnel. - Ben On Feb 13, 2009, at 5:38 PM, Brian J. Murrell wrote:> Am I crazy for thinking about forwarding the DBUS session over an ssh > connection much like the X session is forwarded? > > I've proven the concept with socat (to use the unix socket that dbus > normally listens on rather than reconfiguring dbus to use a tcp > socket) > and it works just as expected. > > Any reason we wouldn't want openssh doing this natively? > > b. > > _______________________________________________ > openssh-unix-dev mailing list > openssh-unix-dev at mindrot.org > https://lists.mindrot.org/mailman/listinfo/openssh-unix-dev
On Sat, 14 Feb 2009 02:20:42 -0600, Ben Lindstrom wrote:> > Any reason why this has to be an unique feature? Is there some form of > authentication cookie or something that needs to be managed?No but there is an environment variable that needs to be set, in the same manner as DISPLAY is set for X forwarding. And the socket being forwarded will most likely be a UNIX (socat calls them "abstract") socket. AFAIK ssh does not handle UNIX sockets currently, does it? It is possible to set the local dbus transport to use TCP (in addition to or instead of) the UNIX socket, but I think making that a pre-requisite for DBUS forwarding is just lame. By default it is a UNIX socket, so SSH should just use that.> That is > really the only reason why X forwarding exists. Otherwise it could also > be yet another port tunnel.How would the setting of the DISPLAY variable be done then? b.
Brian J. Murrell wrote:> And the socket being forwarded will most likely be a UNIX (socat > calls them "abstract") socket. AFAIK ssh does not handle UNIX > sockets currently, does it?I seem to recall previous posts about forwarding UNIX sockets. Maybe it was already implemented by someone? Search archive and bugzilla.> How would the setting of the DISPLAY variable be done then?SendEnv in ssh_config and AcceptEnv in sshd_config can be used to push out environment variables, but I guess you want to customize a little on the server to update the socket name. I guess the objection is simply that dbus is too fluffy for anyone to jump on it, but if you can provide a patch which turns out to be very small and if you can demonstrate/explain how the feature is a great benefit for users then it has a better chance of being included anyway. //Peter