Let's say that you have two computers connected by a network....let's call them Jenn and Ben. They both will have IP addresses: Jenn=192.168.30.1 and Ben=192.168.30.2

Let's say that you are on Jenn and you want to telnet to Ben. From Jenn, you can send packets to Ben by addressing them to 192.168.30.2. But how will Ben know what to do with them when they arrive? This is where port numbers come in.


On Jenn, you would run the telnet command. It will need a local port number and it just pick one. This kind of port is a high number...let's say it picks 5,327. It also looks up it's local IP address which is 192.168.30.1. This pair of numbers (192.168.30.1 and 5,327) will be one endpoint of the connection.

Now it looks up Ben's address which is 192.168.30.2. It knows that it wants to talk to a telnet server on Ben. By convention, a telnet server must use port 23. 23 is a well known port...these are low numbers.

Now the telnet program will establish a connection from (192.168.30.1 and 5,327) to (192.168.30.2 and 23). You need all 4 numbers to identify a connection.

Originally, the pair of numbers (ip address and port) were called a "socket". And it sometimes still is.

But in Unix, socket is also a system call that can be used to write network programs. A socket is also the data structure created by the socket() system. I have heard a few people speak in a way that suggests that they think that a socket is all 4 numbers in a connection. So it's hard to say exactly what a socket is.