During the first twenty-five years of the Arpanet and its successor, the Internet, the Internet was able to take root across the world, prior to booming over the next 25 years to become an unstoppable worldwide phenomenon.
Following French involvement in the initial project development team and the subsequent development of the Cyclades network in France (in symbiosis with the Arpanet), the DGT and computer manufacturers in France chose to pursue different paths. For the latter, their goal was still to develop a model and standards for building heterogeneous computer networks that would enable distributed applications to be run. This implied standards developped through the International Organisation for Standardisation (the ISO).
During that period, I worked for the manufacturer CII-Honeywell-Bull, which became Bull, as head of their network architecture team. I was there until the late 80s, when the Internet re-emerged and gradually began to establish itself in France. As we look at France’s involvement in the rise of the Internet, it’s worth going back over the timeline of events, which many people won’t be familiar with.
Michel Elie is a French engineer specialized in distributed computing systems architecture. He began his career in 1965 as a telematics project manager at the CII. In 1969, as a research assistant at UCLA, he participated in the Arpanet project. On his return to France he worked on the NNA network architecture, adopted by the CII in 1972, and was the technical liaison with the Cyclades project. From 1975 he was in charge of Bull's DSA and ISO/DSA architecture team. In 1988 he became a partner in Bull's Advanced Research and Development department. In 1997, he created the Internet Usage Observatory in Montpellier, France.
Pre 1969 : USA - the potential of computers is explored
Post-war, the US Advanced Research Projects Agency’s network project, subsequently known as the Arpanet, was the first instance of a specific set of ideas, experiences, concepts and potential developments being brought together. Primarily centred around US universities, this is explored in Alexandre Serres’ interesting thesis on the emergence of the Arpanet, as well as Gérard Le Lann's article.
1969 - 1984 : USA - the birth and development of the Arpanet
The Arpanet was launched on 29 October 1969 at the University of California, Los Angeles (UCLA), primarily as a communication network based around packet-switching and the principle of transmitting transparently data from end user to end usr (peer to peer) . The aim was to operate applications distributed between interconnected heterogeneous computers in order to share hardware and software as well as data and human resources .
As a research assistant in UCLA’s computing department in 1969-70, I was part of the network working group (NWG), which included several future “Internet pioneers” : Vint Cerf, Steve Crocker, Jon Postel… and others who might not be so well-known, but who shouldn't be forgotten : Mike Wingfield, Charlie Kline et un peu plus tard Alex McKenzie. The success of such a project hinges as much on the collective intelligence of the team as it does on individual figures, however brilliant they might be.
I was fortunate enough to share an office with Jon Postel, later head of the Internet Assigned Numbers Authority (IANA). As a result, I had a front-row seat when it came to following and discussing the development of network specifications. I proposed a number of minor modifications, in addition to suggesting the development of the Network Interchange Language (NIL). This was subsequently taken up by the Network Control Language (NCL), which I was given the opportunity to present to Michel Monpetit and Alain Profit when they visited UCLA in early 1970.
Before returning to France, I wrote a Master’s thesis, “General purpose computer networks”, with Leonard Kleinrock as my supervisor.
In early 1971, I was the first person to present the Arpanet in France, in a talk I gave to the IRIA and in an article entitled “The ARPA computer network and general computer networks”.
Four years later, the Arpanet had 40 nodes and 45 connected servers. Traffic increased from 1 million packets a day in 1972 to 2.9 million packets a day in September 1973. In 1984, all of the Arpanet’s websites switched to the TCP-IP Internet standard (see Gérard Le Lann’s contribution). There was no difference between the Arpanet and the Internet in terms of their DNA, despite what some would have you believe - they were the same network with the same applications. The only difference was that the new switching platform was more widely accepted, enabling other networks to be connected, with the Internet becoming more of a network of networks. It's similar to the switch from 4G to 5G, which will be invisible to users but which will increase their capabilities.
1972-1977 : France - the Cyclades project takes inspiration from the Arpanet
In 1972, the IRIA launched Cyclades, a French project directed by Louis Pouzin. Like the Arpanet, Cyclades employed packet-switching through the Cigale network. The network’s packet-switching technology was developed on a Mitra 15 minicomputer produced by the CII (Compagnie Internationale pour l’Informatique), a French computer manufacturer created within the framework of the Plan Calcul. I was the technical liaison between the CII and the Cyclades project. Jean Pierre Touchard, CII engineer, is assigned to the development of Cigale's packet switch in Jean Louis Grangé's team. The objective pursued by the CII was to integrate this packet-switching technology into its product range. The first version of Cigale came into operation roughly three years after the Arpanet.
Meanwhile, I was working with Hubert Zimmermann on the transmission architecture model and the end-to-end transport protocol. This work was submitted to the Arpanet’s INWG working group and taken into account during the discussions which led to the TCP protocol.
1971-1989 : the development of the NNA network architecture, coordinated by the successors to the CII
The network architecture adopted by the CII, which was baptised NNA, made it possible to simultaneously manage the datagram and virtual circuit services provided by the “network layer” via the “transport function”, anticipating the standardised OSI architecture model. NNA had been accepted by our partners from Unidata.
In 1974, Unidata was dissolved, and it was announced that the Cyclades network would be shut down. The end of the Cyclades project put a halt in France to the cooperation with the Arpanet team. It stopped the dynamics of users from the universities or public sector enterprises pooled into the Inforep association, in order to promote distributed applications. Ties with the teams working on the Internet in the USA were cut in 1975. At CII-Honeywell-Bull, we didn’t hear any more about the Internet, including from our American partner, until 1985.
When the CII merged with Honeywell Bull in 1976, the expertise of the CII’s engineers in the field of transmissions was recognised. The aim of the new front-end, developed on a mini 6 computer, was to support the Transpac network ahead of its launch in 1977 (see Philippe Picard's article) and to integrate it into the transport layers of CII-Honeywell-Bull’s DSA distributed systems architecture, which was developed between 1976 and 1978 by a team initially made up of CII-Honeywell-Bull and HIS personnel.
1975 - 1985 : the launch of IBM’s SNA architecture and the standardisation of open systems interconnection
Following the emergence of IBM’s SNA (Systems Network Architecture) proprietary architecture in 1975, the primary concern for other manufacturers was to come up with a strategy to prevent the network market being dominated by IBM through its network architecture. They quickly realised that it wasn't feasible for each individual manufacturer to come up with their own network architecture. As a result, they had to choose between :
- Accepting SNA as a universal architecture and manufacturing SNA compatible products, giving IBM the advantage of having sole control over specifications;
- Or coming together to develop a standardised architecture based around the open system concept. In France, this second option was supported by the major users, who called for “open” standards.
Among manufacturers, CII-Honeywell-Bull emerged as leader of the movement for establishing standards for open system interconnection, introducing their own DSA standards into standardisation circuits and delegating experts to attend discussions. In 1977, at CII-Honeywell-Bull’s instigation, the manufacturers created The European Computer Manufacturers Association (ECMA) technical committee 23 on “open systems”. The ISO, meanwhile, created SC 16, the first meeting of which was held in March 1978 and chaired by Charles Bachmann, HIS's DSA director.
1978 also saw the publication of the Nora-Minc report on the computerisation of society: the question of information highways and “networking the world” really came to the fore. The European Commission supported and funded the development of OSI protocols through the ESPRIT (European Scientific Programme for Research in Information Technology) programme, which was launched in 1983. Along with a number of other European ICL manufacturers (Siemens, Phillips, Olivetti, etc.), Bull participated in multiple ESPRIT projects (ROSE, THORN, etc.) in an attempt to accelerate the availability of products which met OSI standards.
1980 to 1989 : France - the DGT seizes the initiative on teleprocessing applications
In 1980, the DGT (Direction Générale des Télécommunications) developed a national videotex system called Minitel, which did not garner much recognition outside of France (the English and German videotex services, Prestel and Bildschirmtext, did not have the same success). Minitel was highly innovative, both in terms of the generalised distribution of terminals for setting up the electronic directory and the kiosk system, which dealt with remunerations for application managers. However, it suffered as a result of the rise of personal computers and its dependence on character mode terminals.
While this was happening, the DGT, at the instigation of Hubert Zimmermann, supported the standardisation of OSI at the IUT, and encouraged its adoption by manufacturers through the Architel project.
Although the switch from the Arpanet to the Internet on 1st January 1983 (“flag day”) had very little visibility in France, the breakup of the Bell system on 1st January 1984 did not go unnoticed, and it was then that the importance of deregulation in the context of the expansion of the Internet began to be understood.
It wasn’t until 1985 that the Internet began to gradually return to Europe, primarily through Unix and the Berkeley Software Distribution, which was provided to universities supporting TCP/IP (but not the character sets, other than ASCII).
But the costly detour via OSI and international standardisation did have some positive effects as well :
- It raised awareness of the issue of networks among a large number of computer scientists and students;
- It helped to establish a conceptual model and a vocabulary for questions surrounding end-to-end information transfer and for teaching about it;
- It helped progress to be made in terms of the language for describing protocols and developing portable “strains” for different operating systems. Bull, which had a very heterogeneous catalogue of computers and operating systems, was thus able to limit the number of developments.
1989 - 1994 : USA - the Internet develops a structure and establishes itself through the Web
Although the Web, which was introduced in 1989 by Tim Berners-Lee, was presented as an Internet application, it brought about a revolution in the sharing of data and the development of hypertext and browsers, such as Mosaïc in 1993.
In France, the decision to opt for a French videotex, Minitel, and its subsequent success slowed down the spread of the Internet in the country. It wasn’t until 1991 that videotex entrepreneurs began to consider investing the substantial profits they had made in the migration towards the Internet (partly through sex-chat services...but not exclusively). Evidence of this can be found in Rafi Haladdjian’s fascinating interview with Valérie Schafer or the success experienced by Xavier Niel.
The Internet was given a structure through the creation of the Internet Engineering Task Force (IETF) in 1986, which helped to ensure its technical development, and the Internet Society (ISOC) in 1992, which placed an emphasis on the perspectives of users.
In 1993, Al Gore, the vice-president of the USA, became the Internet’s number one political supporter, convinced of the how important controlling it would be for the country. The Internet Assigned Numbers Authority (IANA), directed by Jon Postel, was replaced in 1998 by the Internet Corporation for Assigned Names and Numbers (ICANN), a non-profit organisation based in California tasked with the administration of IP addresses and top-level domain names.
This was the USA’s way of making clear their intention to retain control of the Internet rather than delegating it to the IUT, the UN agency for information technology and communication.
1994 to now : France enters the game and the Internet is gradually invaded by the commercial sector
Things got off to an inauspicious start, however. In 1996, France Telecom moved into the Internet market by launching Wanadoo, but continued to prioritise Minitel, despite not making any technical upgrades to it. They had no intention of killing the goose that had laid the golden egg. Up until 2000, manufacturers remained sceptical about the Internet’s economic model. But the detour via videotex through Minitel had prepared the general public for their entry into the digital world, and things quickly picked up speed, becoming exponential. However, a lack of understanding and delayed regulations led to the “Internet bubble” in the early part of this second quarter century.