Donald Davies

Abstract

Donald Watts Davies (1924–2000) worked with Alan Turing on the Pilot ACE at Britain’s National Physical Laboratory, and in 1965 worked out a way to send data through a network in short independent blocks, which he named packets. His colleague’s 1967 presentation of the idea helped persuade Larry Roberts to build the ARPANET on packet switching, and Davies’s own NPL network ran from 1970 into the mid-1980s. In later years he worked on computer security. He arrived at packet switching independently of Paul Baran, and the two men credited each other.

Treorchy, Imperial and the Bomb

Davies was born on 7 June 1924 in Treorchy, a mining town in the Rhondda Valley in Wales, and grew up in Portsmouth, where he went to grammar school. He studied physics at Imperial College London, graduating with first-class honours in 1943, and was then assigned to Tube Alloys, Britain’s atomic bomb project, at the University of Birmingham; his supervisor there was Klaus Fuchs, later exposed as a Soviet spy. After the war he returned to Imperial and took a second first-class degree, in mathematics, in 1947.

The Pilot ACE

In September 1947 Davies joined the National Physical Laboratory (NPL) in Teddington, on the western edge of London, as an assistant in Alan Turing’s group. Turing left for Manchester in 1948 with the ACE unfinished, and Davies was part of the team that finished a reduced version, the Pilot ACE, which ran in 1950 (see Alan Turing and the Enigma). A newsreel filmed at the NPL in July 1949 shows him with a relay machine he had built that played noughts and crosses, showing its moves on a screen; the HNF suggests it may have been one of the first working video games.

He then worked on cryotrons, superconducting switching elements (see Dead End: The Cryotron), and on machine translation, spent a year at MIT and time in India, and in 1963 published a book, Digital Techniques. In 1960 he became head of a division at the NPL.

Packets

In May 1965 Davies went to a computing conference in New York and visited several American universities, where he took a close interest in time-sharing, the new practice of connecting many terminals to one computer. In November 1965 the NPL held a seminar on the subject that brought researchers over from the United States. Time-sharing traffic reminded Davies of the store-and-forward message switching used in telegraph networks, where each node punched an incoming telegram onto paper tape and then sent it on to the next. He applied the idea to data: split each message into short blocks, let every block find its own way through the network, and reassemble them at the destination. He called the blocks packets.

He presented the scheme at the NPL on 18 March 1966 and that month proposed that the British Post Office build a prototype network; in June 1966 he set it out on paper, with packets made of six-byte data words and an assumed line speed of 1.5 megabits per second. The paper cited Paul Baran, who had studied survivable communication networks at the RAND Corporation in the early 1960s and published an eleven-part study by 1964; Baran’s “message blocks” amounted to the same idea, but his work had reached few readers.

Gatlinburg and the ARPANET

In 1967 Davies’s colleague Roger Scantlebury presented the NPL design at the ACM Symposium on Operating System Principles in Gatlinburg, Tennessee. In the audience was Larry Roberts, who was planning a computer network for the US Defense Department’s ARPA. Roberts knew of packet switching, but in the HNF’s account it was Scantlebury’s talk that convinced him of its advantages, and the ARPANET, opened in 1969, was built as a packet-switched network (see Larry Roberts and ARPANET and ARPANET: Building the Network). In June 1972 Davies was one of the founders of the International Network Working Group with Louis Pouzin, Steve Crocker and Peter Kirstein (see Louis Pouzin and the Datagram).

The NPL Network

Davies built his own network too. The Mark I NPL network was developed from 1966 and went into service on the NPL site at the start of 1970; by 1975 it connected two hundred terminals. A Mark II replaced it in 1973, and the network stayed in use until the mid-1980s (the HNF gives 1985, other accounts 1986). Davies wrote up the field in Communication Networks for Computers (1973) and Computer Networks and Their Protocols (1979).

Security

From the late 1970s Davies concentrated on computer security and encryption, and published Security for Computer Networks in 1984. He retired from the NPL that year and kept working as a consultant; MacTutor counts 84 projects between 1984 and 1998. He was appointed CBE in 1983 and elected a Fellow of the Royal Society in 1987.

Davies died on 28 May 2000 in Esher, Surrey, a few miles from Teddington. (The HNF post gives 8 May in its summary and 28 May in its text.)