Two stories that merged
For most of the 1970s, the story of computing and the story of networking ran in parallel. Computers got smaller and cheaper, and networks got larger, but a typical computer user rarely touched a network. By the 1990s the two had effectively merged: a computer without a network connection felt incomplete, and the internet was inconceivable without millions of machines on it. This article follows how that happened through shared rules, spreading software and a widening community of users.
One protocol suite for many machines
The central idea was that machines of any kind could join if they spoke the same protocols. TCP/IP, designed in the 1970s, gave computers a common method for addressing, delivery and error recovery. The ARPANET moved to it at the start of 1983, and soon afterward other services took shape around it. Among the most important were:
- Telnet, for logging in to remote machines, and FTP, for transferring files, both of which predate TCP/IP and were adapted to it.
- SMTP, a standard for sending mail between servers, introduced in the early 1980s.
- The domain name system, developed in the early to mid 1980s, which replaced a single central list of host names with a distributed directory, so that people could type names such as a university’s domain instead of a number.
The technical background of how networks were first connected is in the article on the early days of computer networking, and the design thinking in the one on network architecture. Here the focus is on how the protocols spread.
Unix and the workstation boom
An important driver of the spread was the Unix operating system. A version developed at the University of California, Berkeley, in the early 1980s included a working implementation of TCP/IP, funded in part by government research money, and it was distributed to many universities. Because Unix ran on workstations from companies such as Sun Microsystems, which sold powerful desktop machines to engineers and scientists in the 1980s, networking came built in. A researcher who bought a workstation could plug it into the campus network and be on the internet.
Personal computers followed more slowly. Early PC users were more likely to use modems to dial bulletin board systems, run by hobbyists, or commercial online services. Over time, operating systems for PCs added TCP/IP support, and by the mid-1990s, home users could install software to connect to the internet through a dial-up provider. Connecting an ordinary home computer became a routine task rather than a research project. The history of personal computers explains the machines at the other end.
From academic network to public network
In the mid-1980s, the National Science Foundation funded a backbone linking supercomputer centers and regional academic networks. This network, NSFNET, grew rapidly and became the main route for the internet in the United States. Rules at first limited it to research and education. As commercial networks developed, and as more businesses sought connections, those restrictions were loosened in the early 1990s, and the backbone was handed over to commercial providers in the middle of that decade.
Internet service providers began selling dial-up access to the public in the early 1990s. The number of connected computers rose fast during the decade, although exact counts from this period are unreliable because the network was not centrally registered. The World Wide Web, which arrived around 1990 and spread after the arrival of graphical browsers, gave ordinary users a reason to connect. A distinction worth keeping is that the web is one service among others on the internet, a point covered in the article on web versus internet.
Growing pains
Growth exposed limits in the original design. The number of available addresses, based on a 32-bit scheme, began to look small by the early 1990s, which led to work on a successor, IPv6, and to temporary measures that let many machines share addresses. Routing tables swelled as the network grew, and engineers introduced more flexible address allocation to slow the growth. Congestion, security weaknesses and spam arrived, along with debates about how an open network should be governed. The 1988 Morris worm, discussed in the article on computer viruses, was one early sign that the trust on which the system relied could not be assumed.
Computing changes in response
As networks became the norm, software changed. Programs began to assume a connection: browsers, mail programs, and later software that updated itself over the network. Operating systems added networking components as standard parts, and hardware makers added network ports to their products. Business computing moved toward client-server designs and then to web-based applications. Even the way software was distributed shifted, as downloading began to replace boxes of floppy disks.
At the same time, the internet depended on computing advances: cheaper processors, larger memories and faster modems made it possible for ordinary machines to handle graphics and network traffic. Neither side could have grown alone.
Looking back
Four ingredients stand out. There was a shared set of open protocols that anyone could implement. A bundled implementation in widely used operating systems helped them spread. Government and academic funding built the early backbone, and then a shift to commercial operation widened access. The people who built the web and the first browsers used all of these without needing special permission. That combination explains why, by the end of the 1990s, saying that a computer was “on the internet” no longer needed any explanation.