Louis Pouzin

The son of a sawmill operator, Louis Pouzin followed his mechanical interests to France's top engineering university, the École polytechnique. After graduating in 1952, he designed machine tools for the national telecommunications monopoly. But he found himself drawn to a new technology—computing. After leading a team of software engineers on an ambitious mainframe project at Machines Bull, he moved his family to Massachusetts to work on MIT’s groundbreaking Compatible Time-Sharing System. There he worked on email as well as creating a program called RUNCOM to simplify entering repetitive commands. He called it a “shell,” now a generic term for such programs.

In 1971 he was asked to develop a national computer network for France as part of an ongoing effort to make French computing competitive with IBM. Pouzin toured the US to study the recently deployed ARPANET. He was impressed by its scale but considered its design overly complex and overly reliant on the reliability of the network between the sender and receiver. He also studied the NPL Mark I network developed by British networking pioneer Donald Davies.

Pouzin and his team at the French national computing laboratory named their network CYCLADES, after a chain of Greek islands. Its minimalist design was based around an innovation Pouzin called “datagrams.” Like letters sent by certified mail, datagrams could take any available route and arrive in any order, and if some got lost the receiver would simply ask for them to be re-sent.

The “connectionless” datagram design was ideal for sending data over networks you didn't control. This made it promising for the next huge challenge facing networking pioneers: how to smoothly hook incompatible networks together into networks of networks, termed internets. In 1972 Pouzin was a key founder of the International Packet Network Working Group (INWG), in cooperation with ARPA in the US and two major international standards bodies. The working group assembled networking pioneers from around the world. They had experimental internets running by the mid-1970s, including one that evolved into the global internet we use today.

It wasn't all smooth sailing for Pouzin's datagram approach. He used it for Europe's own experimental internet, the European Informatics Network. But ARPA had broken with the working group's tentative consensus and chosen a connection-oriented approach for its own initial internetting design, Transport Control Protocol (TCP)1.

France's own telecom monopoly also chose a connection-oriented network as the infrastructure for its wildly successful Minitel system, and by 1978 CYCLADES was canceled.

The datagram was down, but not out. When ARPA's TCP internet standard was finally deployed in 1983, it had added a datagram capability. The hybrid was now called TCP/IP, with “IP” standing for “Internet Protocol.”

That compromise didn't mean ARPA had made peace with Pouzin and the other datagram purists themselves. In fact, members of his former team were key players in the official, international internetting standard that became its main rival. ARPA's TCP/IP wouldn't win out as the internet until the late 1980s. Louis Pouzin continues to develop networking protocols, including radical alternatives to TCP/IP and the Domain Name System (.com, .org, .edu etc.). He has received many leading honors for his work on the datagram.

1 TCP was initially tested in 1976 and 1977 from the SRI Packet Radio Research van, now in the CHM collection.