ARPANET

networkingISP & telecom

The packet-switched network funded by ARPA that grew into the internet, beginning with four university nodes in 1969.

Four nodes at the end of 1969 - UCLA, Stanford Research Institute, UC Santa Barbara and the University of Utah - connected by Interface Message Processors, which were the first routers in everything but name: separate machines that spoke the network protocol so the host computers did not have to. That division survives in every network today, where a switch is a computer whose job is to not be your computer's problem. The design decision that mattered was packet switching rather than circuit switching. A telephone call reserves a path end to end for its duration, which is ideal for speech and wasteful for anything bursty, while a packet network forwards self-describing chunks independently and reassembles them at the far end. Everything that follows - congestion control, routing, best-effort delivery, the fact that your connection degrades rather than drops - descends from that choice. Two things commonly said about it are worth correcting. It was not built to survive a nuclear war: Baran's earlier RAND work on survivable communications did address that, but the ARPANET's own motivation was resource sharing among expensive, scarce and incompatible research computers. And it did not become the internet by growing; it became the internet in 1974 when Cerf and Kahn published a way for separate networks to interconnect through gateways, which is a different and larger idea. The network itself made a hard cutover to TCP/IP on 1 January 1983, with no gradual transition, and was decommissioned in 1990 having been made redundant by the thing it started.

Also known as: arpa network, advanced research projects agency network

← All glossary entries