Toby Opferman
http://www.opferman.net
programming@opferman.net


				TCP/IP

INTRODUCTION

	This is a basic tutorial of how TCP/IP works and it will get
you ready for the next tutorial, SOCKETS programming.  


	The idea of TCP/IP networking from a high level point of view
is extremely simple.  One computer has runs a server program.  This
program accepts connections from other computer (Or the same computer)
via a client program.  Now, how does an outside client program reach
the correct server program?  Simple.  First, you need the computers
IP address of course.  Then, you need the port number.  That is it.
The port number can be a range of 0-65535 (Some are reserved for
well known protocols, example you can't use 0-1023 for general use).

	A "port" is just a number.  Here is a simple way to think of
it:


int Hello[65536];

An integer array in C called 'Hello'.  It has 65536 elements.
If the programmer wants put a number in a certain position in the 
array, he needs to know 2 things.  #1, the name of the variable,
"Hello".  And, #2, the element number he wants to put the data at.


Hello[4] = 10;

This is the same with TCP/IP.  The client needs to know the IP and
needs to know the port number.  Very simple.

There are 2 different kinds of connections.  UDP and TCP.
TCP is streaming, meaning there is constant communication from the
client - server and if you send data it will get to the server else
the connection closes.

UDP is not streaming.  It is basically broadcast.  You do not know
if the packet got to the server.   This should be used with any program
that doesn't care wether the information gets there (Streaming audio for example,
if you ever have heard static parts and skipped parts it is because it's UDP. It
is faster to send the data than to send with error detection when you need to be
fast in real time and the information isn't quite that important that loss isn't
a real issue.).
If you do care, use TCP or you'll be forced to reinvent it.


PROTOCOLS

Protocols are nothing more than parcing strings.  You give a defintion of your
protocol, example:

SEND <name> [CNTROL CODE]

You may seperate things by spaces and end each transmission with a control code.
the first word is the "command" of the protocol.  Next comes the data in whatever
format is expected, if sending a binary file you will have to send the # of bytes to count
since your ending [CNTROL CODE] may be apart of the file and try to end your transmission
prematurely.

TALK <message> [CNTROL CODE]

Like the above, make sure your program does not allow the end control code to be sent.
The parcing loop gets the "TALK" command, sends the rest to recieve the message and stops
when it hits the [CNTROL CODE].  You can even maybe have a special \ character instead
where \\ = \ and \[CNTROL CODE] = data control code and not ending control code.


CLIENT/SERVERS

There can only be 1 server per port per computer, but there may be any number
of clients on any computer even the server itself.  It is in your best interest
to make the port configurable so it's easier to set up in case something else is
using a specific port number.

SERVER

To start a server, you create a socket and bind a port to it.  Then, you just
listen for connections from clients.  Once a connection is established, you
perform your interactions and when you are ready you close the connection.


CLIENT

To start a client, you make a socket and you connect to an IP and Port number.
Then once you establish a conection, you perform your interactions and when
you are ready you close the connection.


This was a basic overview on TCP/IP, you should be ready now to learn sockets.