The internet’s fifth man

AT A glitzy ceremony at Buckingham Palace this summer, Queen Elizabeth II honoured five pioneers of computer networking. Four of the men who shared the new £1m ($1.6m) Queen Elizabeth Prize for Engineering are famous: Vint Cerf and Bob Kahn, authors of the protocols that underpin the internet; Tim Berners-Lee, inventor of the world wide web; and Marc Andreessen, creator of the first successful web browser. But the fifth man is less well known. He is Louis Pouzin, a garrulous Frenchman whose contribution to the field is every bit as seminal.

In the early 1970s Mr Pouzin created an innovative data network that linked locations in France, Italy and Britain. Its simplicity and efficiency pointed the way to a network that could connect not just dozens of machines, but millions of them. It captured the imagination of Dr Cerf and Dr Kahn, who included aspects of its design in the protocols that now power the internet. Yet in the late 1970s France’s government withdrew its funding for Mr Pouzin’s project. He watched as the internet swept across the world, ultimately vindicating him and his work. “Recognition has come very, very late for Louis,” says Dr Cerf. “Unfairly so.”

Born in 1931, Mr Pouzin grew up in his father’s sawmill in a village in central France. Drawn to the dangerous machines he was barred from touching—not just the saws, but the steam engine that powered them—he tinkered instead with a Meccano construction kit. His parents encouraged him to apply to the École Polytechnique, France’s most prestigious technical university. After graduating, Mr Pouzin designed machine tools for the state-run post, telegraph and telecoms provider (PTT).

But then, in the 1950s, he read an article in Le Monde, reporting from an annual exhibition of office suppliers at which IBM, an American technology firm, promised that its computers would soon handle all sorts of bureaucratic drudgery. Enchanted by the potential of computerisation, Mr Pouzin moved to Bull, IBM’s French competitor. There he managed a dozen engineers crafting applications for the Gamma 60, a temperamental machine “which filled two large rooms across two different floors”. But the rigours of the job—and Bull’s partnership with RCA, an American company—exposed the limits of his skills. “I realised if I didn’t learn to program or speak English I couldn’t have a career in computing,” he recalls.

A two-year sabbatical at the Massachusetts Institute of Technology gave him the chance to do both. In the early 1960s Mr Pouzin moved his young family to America, where he joined a pioneering team working on time-sharing systems, which aimed to make better use of expensive mainframe computers by enabling several users to run programs on them at once. Mr Pouzin created a program called RUNCOM that helped users automate tedious and repetitive commands. That program, which he described as a “shell” around the computer’s whirring innards, gave inspiration—and a name—to an entire class of software tools, called command-line shells, that still lurk below the surface of modern operating systems.

The French exception

In the late 1960s France’s politicians launched an ambitious plan to bolster the country’s computing industry. In 1971 they challenged IRIA, a state-funded computer-science institute, to begin research into a national computer network. Mr Pouzin was asked to lead the project, which became known as CYCLADES.

Mr Pouzin visited American universities to learn more about ARPANET, a network funded by the military that had been switched on two years before, and which relied on a promising new technique called “packet switching” to deliver data from one machine to another. Chopping up all communications into data packets of fixed size, and allowing machines to relay packets to each other, meant that there was no need for a direct link between every pair of machines on the network. Instead, they could be wired together with relatively few connections, reducing the cost and increasing the resilience of the network. If a network link failed, packets could take a different path.

But to Mr Pouzin, ARPANET seemed over-designed and inefficient. Every computer required a complex piece of hardware to link it to the network, because ARPANET’s design included a connection set-up phase, in which a path across the network was established for communication between two machines. Packets were then delivered in order along this path.

Mr Pouzin’s team came up with a leaner, more efficient way to do things. Instead of deciding in advance which path a series of packets should travel along, they proposed that each packet should be labelled and delivered as an individual message, called a datagram. On ARPANET, strings of packets travelled like carriages of a train, travelling in strict order from one station to another. On CYCLADES, packets were individual cars, each of which could travel independently to its destination. The receiving computer, not the network, would then juggle the packets back into order, and request retransmission of any packets lost in transit.

Such “connectionless” packet-switching reduced the need for sophisticated and costly equipment within the network to establish predetermined routes for packets. The system’s simplicity also made it easier to link up different networks. The first CYCLADES connection, between Paris and Grenoble, debuted in 1973—closely watched by Dr Cerf and Dr Kahn, two American scientists who were by this time mulling how best to overhaul ARPANET. They built on Mr Pouzin’s connectionless, datagram-based approach, so that concepts from CYCLADES found their way into the TCP/IP suite of protocols on which the modern internet now runs.

Connection lost

Yet the innovations that made CYCLADES so compelling to Dr Cerf and Dr Kahn stirred hostility within France’s PTT and other state-run telecoms providers across Europe. Their engineers considered the design untrustworthy and disliked the way CYCLADES removed intelligence from the network. Mr Pouzin did little to calm fears that his network threatened the PTTs’ traditional way of doing things. John Day, an American computer scientist, recalls one particularly fiery presentation in 1976. “Louis showed a picture of a castle, marked ‘PTT’,” he says. “A user hung by a noose from its rampart; others were storming the walls.”

During the 1970s Europe’s state-run telecoms operators were building their own data networks, based on the circuit-switching technology used to carry phone calls. “It was complicated and expensive,” says Mr Pouzin, “and that’s why they liked it.” Georges Pompidou, France’s president, had supported IRIA, but after his death in 1974 the government turned against it. In 1978 the budget for CYCLADES was slashed. “They said, ‘You’ve done a good job. Now go fly a kite’,” says Mr Pouzin.

That same year the PTT switched on TRANSPAC, a connection-oriented data network of its own design. “It was a blunder,” says Mr Pouzin, “a dead end”. It did not seem so at first—TRANSPAC underpinned Minitel, a wildly successful consumer-information service which France’s phone company launched in 1982. Minitel offered French citizens online banking, travel reservations and pornographic chat rooms a decade before the world wide web. By the late 1990s it had 25m users. But it proved unable to compete with the internet and was eventually shut down.

Twenty years after the government junked CYCLADES, Maurice Allègre, Mr Pouzin’s former boss and ally, was still mourning the decision. “We could have been pioneers of the internet,” he wrote in 1999. “Now we are only its users, far from those who decide its future.” Mr Pouzin moved on to other projects and eventually went into academia. “We wasted a lot of a great man,” argues Mr Day. “The French were slow to take to the internet, partly because of this history. But now that the network is a fait accompli, Louis has become their hero.”

“The internet itself has not changed in 30 years. A century from now it must not be the same.”

In 2003 the government named Mr Pouzin Chevalier de la Légion d’Honneur, one of France’s highest awards. Mr Pouzin is now 82, and nominally retired. But like many other networking pioneers he is using his fame to campaign for greater openness and transparency on the internet, as its elegant design comes under growing commercial and political pressure. He is a vocal critic of its haphazard governance, in which key decisions are made by a hotchpotch of companies, charities and well-connected geeks, many of them based in America and largely unaccountable to users elsewhere in the world. He worries in particular about the growing clout of the five or six biggest web firms, which encourage users to stay within “walled gardens” of related sites and apps. To Mr Pouzin, this violates the internet’s tradition of openness. “They’ve recreated Minitel, in a way,” he says.

Around 80% of the new technical standards adopted in recent years were devised by American engineers and companies, he notes. He has lobbied for changes that would make the internet more accessible to non-English-speaking users. That campaign won an important victory in 2009 when ICANN, the unconventional charity which manages the internet’s address system, approved a plan to start issuing domain names (including web addresses) written in Chinese, Arabic and other non-Western scripts.

Despite this decision, ICANN—more formally known as the Internet Corporation for Assigned Names and Numbers—is a particular bugbear of Mr Pouzin’s. Based in California and very loosely accountable to America’s Department of Commerce, the organisation has in recent years worked hard to become more representative of the international community it serves. But some governments would like ICANN’s governance responsibilities—and those of the Internet Engineering Task Force, a loose affiliation of networking experts—transferred to a traditional international organisation such as the International Telecommunication Union (ITU), a dusty UN agency that has long regulated telephony. Handing things over to the bureaucracy of the ITU, however, might slow the development and adoption of new standards. As a result, many countries have concluded that the American-led status quo is the least bad option. Mr Pouzin wonders if forming a new organisation by splitting and combining existing international bodies would be a better approach.

Instinctively an engineer rather than a campaigner, however, his main concern is that the underpinnings of the internet should not become fossilised, but should continue to evolve and improve. “The internet was created as an experimental network,” he says. “It still is one.” He supports researchers in America, Ireland, Spain and elsewhere who are devising ways to make it more efficient and more secure. “The internet itself has not changed in 30 years,” he says. “A century from now it must not be the same.” Mr Pouzin may have helped make the internet what it is today, but that does not mean he wants it to stay that way.