The history of the internet

Birth of the IMP
The initial idea was to link the computers over leased telephone lines, with the computers handling networking on top of their other tasks. But this was replaced by a design developed by Larry Roberts and Wesley Clark, where identical Interface Message Processors (Imps) would form and handle the network, and act as gateways for the Arpa computers.

The problem was that each of the IMPs would need to be connected to all the others, which would become unmanageable as the number of connections grew. The solution had been suggested by electrical engineer Leonard Kleinrock as early as 1961, and independently developed by Paul Baran at the Rand Corporation and Donald Watts Davies at the UK’s National Physical Laboratory.

This became known as a packet-switched network, in which the nodes are connected to just a few others. A message is split into packets, which are sent over the network to their destination one packet at a time.

Each node that receives a packet passes it to a neighbour in the right direction for its destination, and each packet can take a different route from node to node until the collection of packets forming the message have all arrived at the destination, where they are reassembled to recreate the original message.

This has major advantages, in that the nodes can be simple message-passers, packets can bypass faulty nodes and find alternative routes, and failed delivery of one packet requires just that single packet to be retransmitted, rather than the entire message.

This resilience to damage, plus Baran’s position at Rand and his earlier work on warning networks, led to the myth that the Arpanet network was designed for military communications that could survive a nuclear attack.

Arpanet is born
By January 1969 Bolt, Beranek and Newman (BBN) had been contracted to build the Imps that would implement the packet-switching Arpanet. But it was clear that standards had to be established to govern what computers said to each other over the Imp network, which applications would run, and how they would work.

A group of academics representing the sites to be connected formed the Network Working Group and started building standards using the Request for Comments (RFC) mechanism, where proposals were circulated, commented on, refined, recirculated and eventually solidified into an agreed form.

The first result of this was Telnet, which allowed remote users to log on to systems as though they were directly connected via terminals, and the File Transfer Protocol (FTP), which handled file exchange over the network. But a more general system, the Network Control Protocol (NCP), was developed to connect systems symmetrically, rather than in client/server configurations.

The first two Imps were connected in October 1969, and inevitably one of the computers crashed immediately. But the concept had been proved, and over the next two years more systems were connected and the network protocols finalised. By 1971, there were 15 Arpanet nodes connecting 23 systems across the US.

Also in that year, email was reaching maturity. BBN’s Ray Tomlinson had written a mail program that included the ‘@’ symbol to separate names and addresses, and this was adopted, extended and added to the FTP standard for the Arpanet as a whole. By 1973, email made up three quarters of all the traffic on the network.

It was becoming clear that the Arpanet was not the end of the story. It assumed that all the connected systems were of the same kind, and already other packet-switched networks were being developed using different standards and transmission media such as satellite and radio.