Writing a Simple Network Server in Java

Suppose a chat program on your computer needs to exchange messages with a server running on another machine across the Internet. Before any message can move, the program has to answer a question that sounds simple: which computer, out of the enormous number reachable on the network, is the right one? Then it has to answer a second question: which of the many services running on that computer is the one it wants? Writing a network server in Java starts with these two answers, because a server has to be reachable at a known computer and listening on a known port.

This article follows the answers to both questions. It starts with the protocols that carry the data, moves to the addresses and names that identify a computer, and ends with the port number that selects a service and the connection that the two programs finally share.

Think of the chat program as a client and the server as the program on the other machine. The client does not simply announce that it wants to talk to someone. It must specify a computer and a service on that computer. The computer is named by an IP address, or by a domain name that stands in for one. The service is named by a port number, which tells the machine which listening program the connection is aimed at.

Follow the identity step by step. First, the IP address picks out the machine. Second, the port number picks out the service on that machine. Together, those two values identify one specific endpoint. A connection between the client and the server is then identified by both values at once, so the same machine can carry several connections, each with its own port.

Communication over the Internet is built on these two protocols working together. When people say TCP/IP they usually mean the whole arrangement, not one protocol alone. The two halves do different jobs, and keeping them apart in your head makes the rest of the network material easier to follow.

An IPv4 address is a 32-bit number written in dotted decimal form. Dotted decimal means the 32 bits are shown as four decimal numbers separated by dots. Each of the four numbers covers one 8-bit portion of the full address, so each can run from 0 to 255. Reading the address from left to right tells you the four portions in order.

The example address 192.0.2.10 is illustrative and was chosen for this article, not taken from the source. Each of its four parts is a decimal number from 0 to 255, and together the four parts make up the full 32 bits.

IPv6 addresses are larger. Where IPv4 uses 32 bits, IPv6 uses 128 bits, which is four times as many bits. The source describes this size difference but does not walk through IPv6 notation, so this article does not attempt to teach how IPv6 addresses are written.

Numbers are efficient for machines but hard for people to remember. For that reason, many computers can be referred to by a domain name, a human-readable label such as math.hws.edu or www.whitehouse.gov. A domain name can be used instead of an IP address to say which computer you want.

An IP address is enough to reach a machine, but a machine can host many services at once. A server does not simply listen for connections in general. It listens on a particular port. A port number is a 16-bit positive integer, and it selects which service on the machine the connection is aimed at.

So before a client can connect, it must know two things: the Internet address or domain name of the computer where the server runs, and the port number on which that server is listening. Leaving out either one leaves the connection unidentified.