potocols are agreed-upon methods of communication used by computers and, for that matter by people. We have protocols for all sorts of activities. For example, take the protocol for hay ing a meeting. Someone chairs the meeting, states its objectives, decides how long it will last, and then invites people to speak. When each person finishes speaking, control returns to the chair. There are also specific ways for handling special conditions such as interjections ("E, cuse me, but..."), error conditions ("Pardon me?"), and so on. These are all part of the protocol for meetings.
In the computer world, protocols are vital to making communication possible. All sorts of decisions must be made when two or more computers want to send and receive data—for example, which computer should begin the communication, how replies are to be handled, how data will be represented, how error conditions will be handled, and so on.
One of the first considerations in designing the ARPAnet was the need to tolerate unreliability. If a network is to be robust, particularly for military purposes, you can't count on a connection being there. You must assume that although you seem to have sent a message, it might not arrive intact or even at all. This is termed (not surprisingly) an unreliable connection. The communications technology created by the ARPAnet designers to solve this problem was called the Host-to-Host Protocol.
The problem with Host-to-Host Protocol was that it restricted the number of computers that could be on ARPAnet. In 1972, work began on the second generation of network protocols, which gave rise to a collection of protocols called Transmission Control Protocol/Internet Protocol (or the much snappier TCP/IP). By 1983, TCP/IP was the protocol suite for ARPAnet.
"The phrase protocol suite is used to describe a collection of protocols that work together. Usually, the protocols in the suite are built one on top of another. The lowest level of protocol handles the most basic functions, receiving pulses of electricity from the communications medium (usually copper wire but also fiber-optic cable and, occasionally, infrared, microwave, or radio). The next level turns those pulses into characters,and so on until you reach the top layer, which hands data to the application in the size and format it expects. This layering of protocols is the reason protocol suites are also called protocol stacks."
TCP/IP has become one of the most widely used networking protocols. Most vendors of computer systems support TCP/IP in one form or another—and if they don't, a third party will be ready to fill the gap. This means that connecting to the Internet is easy and relatively inexpensive (compared to the proprietary solutions of some vendors).
At the end of the 1970s, other networks sprang into existence. The UUCP network (a loose confederation of first hundreds and now thousands of UNIX machines) was followed in the early 1980s by Bitnet (Because It's Time Network... honestly, but then it was the '80s), CSnet (Computer Science Network), and many others. Some were private (such as CERFnet and Bitnet), some were collaborative (UUCP), and some were government funded (ARPAnet, NSFnet, and CSnet).
Then, like some Nordic saga, ARPAnet begat Milnet (son of ARPAnet), an unclassified DoD network connected to ARPAnet by a gateway. The two networks were called the DARPAnet (catchy, eh?), and eventually this became just the Internet.
Properly speaking, a gateway is a computer that connects two other networks or computers that use different protocols. For example, Bitnet doesn't use TCP/IP, so its protocol needs to be translated before it can communicate with the rest of the Internet. Sometimes the term gateway is mistakenly applied to a bridge, which is a computer connecting two networks that use the same protocols but want to have their message traffic segregated. This is done so that only traffic intended for a destination on the other network is passed; if the networks were just joined together, the combined traffic
could overload them.