Re: Running remote Gnome programs - problems with SSH

[Ian Campbell <ijc25 cam ac uk>]

My understanding is that there should only be one name service running per
session/display. The name server sets a property on the root window which
contains the IOR of the name server.

The reason the remote name service exits is that it sees the property on
the root window and decides it doesn't need to run and so exits.

Other apps which are run on the remote host will look at the root property
to get the IOR for the name service (running on the local host), this
allows them to contact the IOR.

The IOR encodes the host and the port, but I wonder if the fact that you
are tunneling the X session might make some difference.
DISPLAY=remotehost:10 on the remote host (for example), at the local end
DISPLAY=local:0 (or whatever). If you were just using telnet then
DISPLAY=local:0 at both ends. I don't know what difference this would
make.

I'm really just speculating, since I don't know all that much about these
things...

Ian

On Fri, 22 Sep 2000, Geoff Reedy wrote:

> On Fri, Sep 22, 2000 at 05:37:56PM +0200, Graham Leggett <minfrin sharp fm> said
> > Has anyone on this list ever got remote apps to run over SSH ever? Or am
> > I the only one with this problem?
> 
> Here is what appears to be happening.  Applications connect to the name
> server via ORBit (most often with unix sockets).  Therefore, it should be
> enough to run gnome-name-service on the remote host.  However,
> gnome-name-service will only run once for a given display, as evident by the
> following line of output with ltrace 
> syslog(6, "name server was running on display, exiting") = <void>
> 
> Because of this the gnome-name-service started on the remote host will not
> start up because it detects that there is already a name service running on
> the display being used, and the applications running on the remote server
> will not be able to talk to any (local or remote) gnome-name-service.  IMO
> the name service should not look to the display to decide whether to start or
> not since this is not how it communicates (unless I am gravely mistaken).
> Instead couldn't the name service ask the ORB if there was a name service
> registered already and exit if _that_ were true.
> 
> -- 
> Geoffrey Reedy                                       vader21 imsa edu
> 
> "Unix-to-Unix Copy Program;" said PDP-1.  "You will never find a more
> wretched hive of bugs and flamers.  We must be cautious."
> 				-- DECWARS
>