The way we accomplished this feature was to setup in the firewall a link
between the client and the server.  We allowed use of all UDP and TCP ports
but it could only take place between those 2 IPs.  When we first attermpted
this, we ran into the problem of the server wanting to tell the client a
different port to use each time.  The ports were random so if you ran some
traces you could porbably narrow down a range, but we found it easier to
just allow those 2 IPs to use all the ports and not have to worry about
figuring which ports the servers was randomly choosing.

If you want to try to configure the firewall by opening ports then I know
you will have to open 1500 and 1501 TCP.  Of course all of this is going to
depend on which side of the firewall your server and client fall behind.
In our configuration the client was behind the wall, so the servers packets
were getting sent back however the client wasn't able to send anything back
because the firewall wouldn't let him.

In my opinion I think just establishing in the firewall a link between just
those 2 IP address is the easiest way to go about things unless you just
like a challenge.