[omniORB] Problems with more than one network interfaces

[Duncan Grisby]

On Friday 16 November, "Hinkofer, Fa. pds, E230, D" wrote:

> We communicate via two network interfaces on each plattform . One is
> a token-ring interface and one is an ethernet interface. Also we are
> using omniNames for the naming service. omniNames is running on
> WINNT4.0 (intel).  And we communicate via the ethernet interface to
> omniNames.
>
> Now the problem: It's possible to register our Objects in the naming
> service (omniNames). But it's not possible to communicate to the
> Objects by using the entries of the naming service. If we use only
> one network interface than we can communicate to the Objects by
> using the entries of the naming service. Is it possible to use two
> network interfaces with omniNames280 ? If it's possible, what do we
> have to do to use the omniNames to communicate to our Objects by
> using the entries?

I don't completely understand what you are trying to do. You can force
omniORB servers to put a particular interface address in its object
references using the -BOAiiop_name_port command line argument or by
setting the OMNIORB_USEHOSTNAME environment variable. That way, you
can force the use of the Ethernet interface.

What you can't do with omniORB 2.8 (or 3.0) is publish both interfaces
in the object references, and have clients connect to the one that
works.

> In the moment it's not possible for us to change the version of
> omniORB.  With the next roll-out of our system we would like to
> change to a higher version of omniORB. Whitch one would like to be
> the best version to resolve our problem ?

With omniORB 4, servers can publish multiple addresses in IORs, and
clients can be configured to pick out the one that is suitable for
them, so that would be the best option.

Cheers,

Duncan.

-- 
 -- Duncan Grisby  \  Research Engineer  --
  -- AT&T Laboratories Cambridge          --
   -- http://www.uk.research.att.com/~dpg1 --