III - Packet Switching Networks
Early research efforts to develop a way of transporting computer data
led to the development of what is called "packet switching". Packet switching technology breaks a message into small sections of data, gives each of these addressing information called a header, which together with the data are called "packets". It then routes and delivers the packets, interspersed with other packets from other messages. After the packets reach their destination, the message is reconstructed. Paul Baran in the US and a few years later, and unaware of Baran's work, Donald Davies in the UK, developed similar concepts. In 1966 Davies implemented a packet switch connecting a set of host computers. Paal Spilling, a Norwegian Internet pioneer, refers to the resulting National Physical Laboratory (NPL) network as the first packet switching local area hub network. (Spilling, Email, August 2002)
In the US, there was interest in exploring the feasibility of packet switching for resource sharing computer networks. This interest led the Advanced Research Projects Agency (ARPA) to recruit Larry Roberts, a researcher at MIT's Lincoln Laboratories to join the Information Processing Techniques Office (IPTO). IPTO was planning to establish a packet switching network interconnecting a number of geographically dispersed dissimilar computers.
Networking technology was also of interest to other researchers around the world. In the early 1970s in France, Louis Pouzin was developing a French packet switching network, building on the lessons learned from previous packet switching research. He studied the research developments in the US and Great Britain, and along with his research group, created the Cyclades/Cigale network. In the UK, the NPL network was being developed by a research group headed by Donald Davies. In the US, there was the ARPANET development.(6) The question became how could these networks be interconnected, i.e. how would communication be possible across the boundaries of these dissimilar networks. (Ronda Hauben, “The Birth of the Internet”)
A plan at the time was to connect the ARPANET in the US, CYCLADES, in France and NPL in Great Britain. A memo written in 1973 describing early technical plans for this interconnection, included a diagram of these three networks linked by gateways. These gateways would make it possible to transmit messages across the boundaries of different constituent networks. Following is a replica of the diagram (Cerf, Memo, p. 5. See Also Graphic I):
(Host)
/
/
( ) ( ) ( )
( ) ( ) ( )
( ) ( ) ( )
(Host)---(CYCLADES)----(gateway)---( ARPA )-----(gateway)----( NPL )
( ) ( ) ( )
( ) ( ) ( )
( ) ( ) ( )
\
\
\
(Host)
Also there was a diagram of data going from a host computer on one computer network to a gateway and then to a host on another computer network.
(H)--(G)----(G)---(H)
\ /
\ /
(H)
Another description of the goal of connecting these 3 different networks, is presented at the International Institute for Applied Systems Analysis (IIASA)in Laxenburg, Austria, in 1974. In a paper for a conference there, British researcher, Donald Davies writes:
“To achieve...the interconnection of packet switching systems we have to decide at what level they will interwork. The levels chosen
could be character stream, packet transport or the virtual circuit.
After some discussion, a group including ARPA, NPL, and CYCLADES is
trying out a scheme of interconnection based on a packet transport
network with an agreed protocol for message transport....
(Davies, “The Future of Computer Networks”, IIASA Conference
on Computer Communications Networks, October 21-25, 1974, p. 36)
Davies’ paper is helpful in documenting the interest in creating a metanetwork of other networks including the ARPANET, NPL and CYCLADES. Also, however, the occasion of the paper is significant. The IIASA is a research institute which supported collaboration among researchers from the Soviet Union and Eastern European countries and from the US, Western Europe and Japan. The conference in 1974 at which Davies spoke was a conference where researchers from these different countries were all introduced to networking technologies and developments of the time, including the ARPANET, NPL and CYCLADES developments.
At a workshop the following year in Laxenburg, in 1975, sponsored jointly by the IIASA and also the International Federation of Information Processing Organizations (IFIP), another British researcher, Peter Kirstein presented a paper that described the collaboration between the UK and the US in networking. The paper included a diagram of the satellite and ground connectivity between the ARPANET in the US and the University College London computers in UK. The diagram also showed the Norwegian connection to the US and UK networks. Kirstein’s paper, “The Uses of the ARPA Network via the University College London Node” was reported to have been exciting to those present and plans for a network connecting the researchers of the IIASA were developed. The list of those at this workshop included researchers from Austria, Belgium, France, the Federal Republic of Germany, the German Democratic Republic, Hungary, Italy, Netherlands, Poland, Switzerland, the Soviet Union, the UK, and the US. Davies and Kirstein were there from the UK, Cerf from the US, Lazzori from Italy. Kopetz from Austria, K. Fuchs-Kittowski from the Germany Democratic Republic. Also there was discussion at the workshop about what kind of network researchers the IIASA would develop to support their collaboration.
The IIASA conference in 1974 and the workshop in 1975 include reports on the networking research being done to create the Internet and other networks like the EIN. It is significant that at a group including researchers from both Eastern Europe, the US and Western Europe, the details of the internetworking developments were presented and discussed. Fuchs-Kittowski, a researcher from the GDR present at the 1975 workshop, remembers discussing possible participation in the UCL network in the UK by those from the German Democratic Republic. (See for example, Graphic III) There is at least one discussion in 1976 about whether or not to have an IIASA connection to the ARPANET or to the European Informatics Network (EIN). There was also international collaboration as part of the IFIP 6.1 working group toward the development of the Internet.
There are various streams of research that made contributions to the development of the Internet. The French research developing Cyclades/Cigale is an important example. The French contributed the concept of the datagram as a means of transporting data. Pouzin also is credited with the creation of the sliding window as a flow control mechanism.(6) There were discussions among those participating in the INWG, later called IFIP, WG 6.1, where decisions were considered about what the standards should be to create the protocol for an Internet. For example, Pouzin describes some of the meetings:
“Within INWG, which joined IFIP as WG 6.1, we had lengthy discussions about which level of protocol should be agreed first. It must have been during an INWG meeting on a boat (Stockholm-Turku and return) that a consensus developed on the principle of a common packet format. I don't have a record of this meeting in my diary, but I gather it was in August 1974, at the time of an IFIP Congress.”
(Pouzin, Email, April 28, 2003)
Pouzin was also at the IIASA 1974 conference and describes some of the discussion there. He writes:
“Yes, this was 21-24 October 1974. We kept refining a common packet format. I had cranked up a proposal overnight during the workshop, and I remember Peter Kirstein made some objections after a call to Vint Cerf in the US. I don't know if this paper was recorded in history, perhaps as an INWG note.”
(Pouzin, Email, April 28, 2003)
Describing the efforts that were made to link Cyclades and NPL, Pouzin explains:
“In the end, there never was an interconnection based on this plan. What occurred was a demo during an ICCC conference in Toronto, 3-5 August 1976. There was a Cyclades terminal concentrator (like a TIP) connected to Paris with a leased phone line. There, a link to NPL was using the packet network EIN (alias Cost 11), (if I remember). Then at NPL it was connected to the internal local net.
On the exhibit in Toronto, Derek Barber demonstrated using an NPL host through this patchwork. I felt it was amazing, if rather intricate.
Another more elaborate attempt was the definition of a protocol subset allowing a TCP-IP host to talk to a Cyclades host, without a gateway, simply by using a restricted set of protocol features. This work was carried out by Alex McKenzie from BBN. He wrote an INWG note. Maybe someone has a copy ! Presumably, there was not enough steam, and money, to implement the idea.”
(Pouzin, Email, April 28, 2003)
|
