Re: panel error *** ORBit patch ***



On Thu, 13 Aug 1998, James Michael Mastros wrote:

> > Both panels have a UNIX socket path of /tmp/username/GNOME-panel, but
> > since they are on two different hosts, there is no namespace clash.
> Hmm?  Why?  Shouldn't that be /tmp/username/IOR:<ior_of_panel_on_A>, and
> /tmp/username/IOR:<ior_of_panel_on_b>, where ior_of_panel_on_A !=
> ior_of_panel_on_B, and another file /tmp/username/orbit?  The two would have
> different IORs, so there is REALLY no namespace clash.

If you want to use UNIX sockets, to get an IOR requires a UNIX socket
path.

With your idea, to get a UNIX socket path would require an IOR. 

Nice loop :)

> So, the sst process, which is on host B, checks if ORBit is already
> running.  It sees /tmp/username/orbit, and knows that it is.  It asks
> for the IOR of panel on A, and ORBit (it is an object request broker,
> after all) notices that it isn't host A, and finds host A and gets the
> IOR?  That dosn't require any persistincy of IPs... mearly the ORBit on
> B trying to send a packet to itself using the given hostname (or IP) of
> B, and not reciving it (and not getting a ENOSUP or such).  If you get
> the packet back, then host A == host B.  If you don't, then host A and B
> are distinct.  If you had network problems, then you should have gotten
> an error, and not a packet that simply went off into the void. 

. There is no program called "ORBit" that you can run. ORBit (and the ORB
  proper in other CORBA packages) is just a library.
. You may be thinking of the naming service, which acts roughly like
  DNS for CORBA objects (give it a name, it gives you an object
  reference if a running object has registered itself under that name)
. If you're not thinking of the naming service, what are you thinking of?
. If you are:
	. The naming service is just another application program, and does
	  not know anything more about which host an object is on than any
	  other application. The location of objects is hidden down inside
	  the lower levels of the ORB, as it should be - CORBA is supposed
	  work transparently on objects in all locations.
	. The naming service would be run on a per-desktop-session basis,
	  and not on a per-host or even per-user-per-host basis.
	. From the two above points, it should be fairly obvious that "the
	  naming service on B" cannot just "find host A and get the IOR
	  for the panel on A"...

. Your approach for finding out whether an address is on the same host
  as a program is pretty cool!

> OK, so tell me why I'm crazy.

Hope this helps ;-]
-- Elliot
Progress (n.): The process through which Usenet has evolved from smart
people in front of dumb terminals to dumb people in front of smart
terminals.  -- obs@burnout.demon.co.uk





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]