A Technical History of CYCLADES

Louis Pouzin Interview 

The following is translation of an interview of Louis Pouzin conducted by Alain Simeray with Autrans'98. The actual interview (in French) can be found at The AUTRANS98 Website.

Louis Pouzin, today retired, is the "father" of the Cyclades network. A network implementing packet switching, such as current used in IP. Today, he has a tooth against Microsoft.

Autrans, 10/01/98

Alain Simeray: Why doesn't anyone speak of the Cyclades anymore?

Louis Pouzin: The first demonstration of Cyclades took place in 1972. The network was operational in 1974. It was extint starting from 1978, fault of financing. Indee, after the phase of research and experimentation, the financing must come from the industry. That was not the case.

AS: What was the network used for?

LP: Originally, the Information Delegation wished to promote French industry, in particular, to inter-connect the data bases of great French administrations. The data bases were everything at the time -there and each one had it's own format. Cyclades was to carry out the intercorrection and contribute to avoi the inconsistencies. We succeeded well in carrying out the interconnection, but not to avoid inconsistencies, which was not our forte.

Cyclades was an operational network which the company CII inspired in order to create products and protocols. The network allowed exchanges of files and batch remote. Thus, for example, during three months, the computer IRIS 80 of the IRIA was broken down. It had heated and was full of dust and it had to be cleaned, each component of it. During this time, Cyclades allowed sending of batch processed for execution on the IRIS 80 on the site Grenoble. One could also use electronic mail, more rustic than that which one knows today of course.

AS: It is said that Cyclades is based on the sames architecture as the Internet...

LP: Completely. Cyclades is based on packet-switching, like the Internet. In fact, all started with a report/ratio of Paul Baran (Rand Corportation) who, in 1963, described a reliable network using the communication of packets independent of one another. In 1967, Larry Roberts, who animated the ARPA, launched the project of a computer network allowing to share resources, thus motivating the researchers who could work remotely on their computer, by preserving their usual environment.

AS: And the fear of a nuclear strike, destroying the system of defense?

LP: You know, to agitate an atomic threat, in full cold war is a good means of obtaining financing on behalf of the congress.. twenty million dollars. Let's return to the History. The first document on the ARPA network was thus published in 1968. Thus, basic architecture was known as of this date, on paper, because the demonstration had yet to take palce. At the beginning of the Cyclades project, I went to the United States and was consulting the company BBN, the same one which working with the ARPA project. The ARPA network implemented packages in virtual circuit, our network of packets was "pure", with independent datagrams, as in IP today. IP took from the Cyclades the "pure" packet and logical addressing and not the physical addressing which was used by ARPANET. I had also introduced into Cyclades the concept of zone, which one calls field in the Internet, as well as the mulihoming and the concept of window slipping for the control of flow. TCP/IP is thus an improved version of Cyclades. It adds in particular the possibility of splitting up the packets in the course of route, a little like a train whose wheel base varies according to networks. This concept is very useful to correct local area networks Ethernet, which did not exist when we conceived Cyclades

AS: How do yo consider the future of the Internet?

LP: I think that the operators (France Telecom, British Telecom..) will take the control of the transmission. In spite of the possibilites of transmission by cable, satellite, radio, it is extremely probable that the majority of the transmission will be held in the telephone network. Indeed the telegraphic networks with ADSL technology offer flows sufficient for the residential user, up to 6Mbps. For 99% of people, it is suitable. The cable is a shared network. Thus a flow of 4 Mbps is theoretical for a user. Put on thousands, they share the flow. The satalitte functions differently. It requires a more expensive infrastructure, that parabola, the decoder... Moreover, to announce flows of the command of Megabit/second is about propaganda. One should rather speak about a few hundreds of Kilobits per second. It is very useful to connect isolated places. Moreover, the operators have the capacity to carry out investments in the long term and they have the means and they are owners of the networks. Lasty, they can manage the subscribers. These elements are significant, because it will be necessary to associate a quality of service that only the large operators will be able to propose, from beginning to end network.

AS: You made yesterday an intervention crashing to pieces at the time of the round table. You have a tooth against Microsoft. Why?

LP: I do not have anything against Microsoft a priori, except that it acts as a quasi-monopoly which destroys the competition. In the field of networks, compatibility is a significant element. However, Microsoft permanently changes version of protocol so that a concurrent expertise cannot develop. It is not normal that world communication systems can develop without being controlled. Moreover, Microsoft creates for itself a protected market. It finances products which it diffuses free in order to kill competetion, and then the fact of paying, to finance a new generation of free products... It is dumping, a reprehensible practice. This is way I speak about "racket". To kill its neighbor, even in an economic way, does not form part of the capitalist model.