The word protocol entered networking in a 1967 British memo
Computers can only talk if they agree beforehand on exactly how. Those agreements, covering message formats, meanings, timing and what to do when things go wrong, are called protocols, a term first used this way in April 1967 at Britain's National Physical Laboratory. Your email, web browsing and Bluetooth headphones all depend on them.
A protocol is to communication roughly what a programming language or an algorithm is to computation. It fixes the format of each message and the meaning intended to trigger one of a set of expected replies, independent of the hardware or software that carries it out. Because all sides must agree, protocols often become formal standards, set by bodies such as the Internet Engineering Task Force and the World Wide Web Consortium for the internet, the IEEE for Ethernet, and the ITU-T for the telephone network.
The 1967 memo, written by Roger Scantlebury and Keith Bartlett under Donald Davies, the pioneer of packet switching, described rules for the NPL network. On the American ARPANET, Bob Kahn's 1822 protocol handled host messages from 1969, and a Network Control Program built by Steve Crocker and fellow graduate students, among them Jon Postel, followed in 1970. In France, Louis Pouzin's CYCLADES network made the computers at each end, not the network, responsible for reliable delivery. Kahn and Vint Cerf then devised the Transmission Control Program; Cerf wrote its specification with Yogen Dalal and Carl Sunshine in December 1974.
Early networking software was one big block, but borrowing from compiler and operating system design, engineers split it into cooperating layers, each handling one class of problem such as transport or addressing. TCP was reworked into the layered TCP/IP stack, adopted on the ARPANET in January 1983. The rival OSI reference model appeared in 1984, and for some years in the late 1980s and early 1990s engineers and even governments argued over which approach would win.
Protocols encode messages either as readable text or as compact binary. FTP, SMTP for email and early HTTP send plain lines of text that humans can inspect while debugging, whereas HTTP/2 and HTTP/3 use binary for speed. A typical message has a header carrying addresses and control fields, and a payload carrying the content, with long messages chopped to fit the network's maximum size.
Source: Communication protocol
