I'm actually trying to do the same thing (write a simple instant messenger program) but am running into a few problems.

Right now I am just focusing on the "communication" portion, using a command-line interface. I plan to add the GUI stuff later.

I am using C++ and the sockets api and have taken the approach of building both a "Server" and "Client" class. I have designed my instant messaging application to have both the server and client in one executable program.

So, in my main() function, I will:
1) Create an instance of Server ( for example, "Server my_serv( ); " )
2) Create an instance of Client ( for example, "Client my_cli( ); " )
3) Start the server ( for example, my_serv.start() )
4) Client stuff... ( connect to remote computer, enter text data to send, etc ).

One problem I have run into with this approach:

I attempt to start the server then use the client functionality, but then my program hangs, because the socket program is listening for a connection. ( I've tried non-blocking IO with the fcntl call below ).
This bit just gets all current flags for the file descriptor sd1, then adds the O_NONBLOCK flag to it.

Code:

  rc |= fcntl( sd1 , F_SETFL , ((fcntl( sd1 , F_GETFL ))|O_NONBLOCK) );

I believe I might need to use the select() function, but I'm not sure if I'm headed down the wrong path.

Would it be best to split this up into 2 programs ( i.e. one as a server and the other as a client )? Or could (should?) I use fork(), start the server in one process, then the client in the other? (I've tried the fork()ing approach, but had some problems.)

I'm trying to get an idea of how the program would be best designed. Any ideas are appreciated.