On Wed, 1 May 2002, Bob Tanner wrote:
> > With portSwap: on the client actively opens the connection to the server. 
> > This is enough to get through masqerading hosts, as well as firewalls wich 
> > are open for connections from the inside.
> I assume portSwap: is a cow only parameter, based on this post from Alec?

Paradise 2000 supports the UDP portswap option, but I used the name
udpTryPortswap for the parameter.  It can also be changed while playing from
the UDP menu.

> I'm assuming you can config the client to use something like port 9999 with the
> portSwap feature.

Either with the baseUdpLocalPort option or the -U command line parameter.  But
this often isn't necessary, as most NAT setups will automatically allow
incoming udp traffic once outgoing traffic has clued them into the
connection.  That's the whole point of portswap after all.

> My confusion is with James posting here:
> http://archives2.real-time.com/vanilla-clients/2000/Mar/msg00008.html
> His recommendation is to use trekhopd, is that still the best recommendation?

I don't think James is seriously recommending using trekhopd.  trekhopd is a
major pain to use, as you need to compile a special client just for it.  It
also doesn't apply to the way most modern filewalls work.  It's really
designed for the "one unix machine with internet access" model that was more
common 10 years ago.  Just forget trekhopd.