Thread: clent-client communication inc

No difference, really. You just have to move the data the right way:

1) Client A sends data to Server, Server relays data to Client B. Client A has effectively sent Client B data. (Probably the easiest, but involves data being transferred twice. Example: Chat)

2) Client A sends Server a request to tell Client B that A wants to connect to B. Server relays the request to B, who then accepts or denies it, and B opens a port for A. Server sends Client A Client B's IP address/port. Client A connects directly to Client B, and data transfer commences. (More complex, involves two connections, possible firewall issues, but data doesn't flow throw a central server, easing load. Example: Filesending in AIM, etc.)

if you don't want a main server to do things for you, like Cactus_Hugger suggested in option #2 (btw, he is right), you will need to think a little different;

Instead of two clients, how about two servers?
Each "server" will be listening for a connection on a defined open port, let's say 666.
So once side A will want to talk to side B, he will send packets to port 666 to IP of A, and the other way around. So both sides listen on the same port and send packets to the same port on the other side.

This will create a pretty flexible stand alone clients, the only problem you might have here, is how will you know what is the IP of each "client/server" machine.