[techtalk] gnome inter-panel rivalry

[Conor Daly]

or "strange behaviour of exported gnome panel"

curiouser and curiouser...

I've been playing around with gnome panel lately and have noticed a few
"features".  System layout...

_____________
|           |
|  Server   |
| hobbiton  |
|  RH 6.2   |
|           |
-------------
      |
      |______________
      |             |
______|______  _____|______
|           |  |          |
|  Client   |  |  Client  |
| valkerie  |  | albatros |
|  RH 6.2   |  |  Win 98  |
|           |  |          |
-------------  ------------

I have a nicely layed out panel set up on the server (we'll call this
panel-h).  On the RH client I have the local panel running (panel-v).  On
the Win98 client I use an X-server.

For the Win98 client: 
I export the display and run panel.  Everything works nicely but if I make 
changes to the panel itself, these changes are lost on logout.  This 
happens either when I use the panel "logout" command or if i allow the 
Windows X-server to kill the panel.  

If I start a term, su to another user, export their display and run panel 
I get told "There is a panel already running.  Do you wish to start another 
one?".  This is fine and is expected behaviour.

For the RH Client:
I export the display and run panel.  Now stuff *really* begins to happen...  
Remember, the RH client is already running a panel (referred to as panel-v) 
and we're now running a panel from the server (called panel-h).

1). When I start panel-h, I get told "there is a panel already
running..." (panel-v I guess).  Now, which panel has detected this and is
asking about it.  I presume panel-h knows nothing about the client host
except that it is showing on a remote display (It probably doesn't even
know this; it has no need to know) so that leaves panel-v or the gnome WM
itself.  Could be a situation like an ethernet card broadcasting its IP
address and another machine responding that it is already using that IP.
Why should panel-v/gnome know/care that there is a panel from another
host running also?

2). panel-h contains a couple of applets (clock and mailcheck since the
server is the mail server also).  When panel-h collides with panel-v,
these applets dissappear from panel-h and appear on panel-v so I end up
with 2 clocks in panel-v and none in panel-h and the mailcheck also
migrates to panel-v where it happily monitors the mail spool on the RH
client which never gets any mail since that all lives on the IMAP server!
If I then logout from the RH client and login at the server's console and
startx, panel-h has no lost its clock and mailcheck!  This behaviour seems 
bug-like to me. 

Anyone come across similar?

-- 
Conor Daly <conor.daly at oceanfree.net>

Domestic Sysadmin :-)