By: Yngvar Lundh, life member IEEE, professor emeritus University of Oslo.
The technology underlying the Internet was developed in the 1970s by a handful of academic groups, in an admirable cooperation. I accepted an invitation to join in the development work, following the impressive demonstration of resource sharing in Washington, DC in October 1972. At the time that goal was especially attractive on the background of computer cost versus the promising visions for computing. As it turned out, the possibilities envisioned were not only fulfilled. They grew and expanded, and are probably still only partly exploited, despite the unbelievable development we have seen so far.
Arpanet – a networking laboratory.
In my work as a research engineer at the Norwegian Defence Research Establishment – NDRE – at Kjeller, near Oslo, Norway, I had been collaborating with ARPA for a few years in the mid 1960s about establishing a seismic observatory in Scandinavia, an effort towards international control of underground nuclear tests. In 1972 we had a leased line between the Norsar observatory at Kjeller and a collaborating seismic site in the US. That transatlantic data line, already paid for, provided an opportunity to establish, within budget, a permanent connection to the new and exciting “Arpanet”: The 2.4 kb/s modems were replaced with newer ones that augmented the capacity of the line to 9.6 kb/s. Multiplexers made the extra 7.2 kb/s available to us as an independent line while leaving the seismic collaboration essentially unchanged. A new node in the Arpanet was established by a “TIP” at Kjeller in June 1973. Soon thereafter a second European node was established at the University College London – UCL – using a leased line from Kjeller. The next European node was established in Germany in 1980.Each node of the Arpanet consisted of a dedicated computer, programmed as “Interface Message Processor” – IMP, with line connection to one or preferably more other IMPs and interfacing to one or more host computers. An extended version of the node computer was “Terminal IMP” – TIP.
Since the first four nodes of the network were installed in 1969 the network had grown fast. The company Bolt Beranek and Newman – BBN in Cambridge, Mass. prepared and installed the nodes. In 1974 Arpanet extended from Honolulu to Kjeller and London. Most of the lines were leased lines transmitting 56kb/s. Several computer networks already existed in the 1960s, associated with the leading computer manufacturers. Arpanet had some unusual features. First of all it was built on a set of modular communication protocols designed to allow computers of any manufacture and any operating system, rather than being restricted to those of one company. Secondly, and very im-portant, it had a network control center – NCC – at BBN. That permitted all traffic to be observed, hence traffic performance to be studied in detail. Each node computer (IMP or TIP) could be repro-grammed from NCC. Thus the network was a very powerful laboratory. And it allowed collaboration over long distances essentially without need for travel. Communication techniques could be studied experimentally. All information was transmitted between the nodes in “packets”. That is limited length strings of binary coded information augmented by head and tail with additional information allowing packets to be handled by the nodes for transfer through the network to destination similarly to packets in the post. Longer strings are chopped up into suitable packet size. Packet handling happened according to the communications protocols that actually were under further development themselves. This novel transmission method, called packet switching, provides great flexibility and efficient exploitation of network transmission capacity while sharing between multiple users. “Protocols” are forms of mutual behavior and expression suitable for interaction between parties in separate environments that may be different. Protocols are useful between tasks in different computers as well as between diplomats from different cultures.
The Arpanet was a powerful laboratory in two respects. It facilitated development of the general communication protocols for computer networking, and experiments with computer networking for resource sharing. Those resources comprised not only the computing power, but many more such as data, personnel, ideas, special software, and others.
The conference in Washington in October 1972 demonstrated many aspects of networking by pre-senting some early collaboration projects. As an example, one computer working on natural language understanding collaborated with another one working on weather forecasting. You could ask the first one a question such as “how is it New York”. That system would find the meaning of your question and then contact the collaborating computer somewhere else to find out what the weather was in New York, and return an answer to be presented in natural language. A long list of rather unusual collaboration projects was demonstrated. It did indeed inspire me to collaborate.
A whole new world of networking and a new network culture were emerging in those years. Enthusiasm was great in academically inclined places. The network grew fast. Exciting ideas proliferated. In spite of ARPA being part of the Department of Defense no military issues or military resources as such were part of the project. The work was aiming steadily all the way for the goals outlined early in the project, although specified more clearly along the way as we gradually understood the issues better. It was all unclassified and open.
Basic technical research
Advanced Research Projects Agency - ARPA was founded by President Eisenhower in response to the request for more basic research spawned by the Soviet launch of Sputnik in 1957. ARPA began to nurse the idea of computer networking in the early sixties. Lawrence (Larry) Roberts, then at MIT answered to a request for proposals. His ideas earned him an employment with ARPA to supervise the building of an experimental net. BBN was awarded a contract. One of their engineers Robert (Bob) Kahn had a leading technical role in the project there. Several networks were growing around the world for various commercial purposes in the 1960s. And much creative work on computer networks went on in the academic world. Douglas Engelbart showed an impressive set of his own and other ideas in a celebrated demonstration at Stanford University in 1968. It appeared as an answer to a challenge made by Vannevar Bush in 1945: “In the future machines will perform much brainwork and communication similarly to how machines are already augmenting human hands”. Engelbart called his project Computer Augmentation of the Human Intellect. It showed many, if not most of the important tools and features that we have become accustomed to while using Internet now. They were rather exotic then. Let me mention some: Workstations in network, CRT screen for text and images, mouse, hypertext.
The first four nodes of a network called Arpanet were installed in the fall of 1969 at UCLA, UC Santa Barbara, University of Utah and Stanford Research International. A set of basic protocols had been devised. Useful services were electronic mail, file transfer, and “telnet” – a protocol that allowed typewriter terminals access through the network to remote host computers. Its importance should be seen in the light of time sharing access to large expensive computers. Personal computers were many years away.
All information transfer between the nodes went in packets. “Packet switching” was a technique that had been studied by Leonard Kleinrock in his PhD project at MIT. His thesis was issued as a book later on.
The fertility for new ideas in that networking environment was amazing. That was amply shown in the exhibition that Bob Kahn organized in Washington. The success inspired Larry Roberts and Bob Kahn to explore the possibilities for an international network. They visited Norway prior to the Washington exhibition.
Ambitious goals and careful development
The network had many shortcomings then, compared to what has become of it eventually. Several researchers began meeting from time to time for discussions, and many more mature ideas evolved. The groups from various academic places called themselves together “Packet Switching Project Working Group” – PSPWG. I remember especially a two-day meeting on board the large ferry between Stockholm, Sweden and Aabo, Finland in late 1974. The PSPWG and a number of researchers from several countries, notably from France and Sweden were there in addition to the American participants from PSPWG. Peter Kirstein from University College London and I came. Creative discussions went on during those two days between some twenty engineering enthusiasts. While sacrificing the beautiful views through the Aaland archipelago participants engaged in technical discussions. They basically covered much of the grounds on which a computer network should be built. I remember the ultimate goals as: “All” kinds of computers should be able to communicate, the needs of a set of specific service types should be met optimally, all kinds of information carrying media having different capacity and characteristics (cables, leased lines, radio, etc) should be optimally usable, and the network should be robust and function without any centralized control. Many of the protocol details for such a net were described in an article in the IEEE Proceedings in May 1974 by Bob Kahn and Vinton (Vint) Cerf. Bob Kahn was employed by ARPA. He took over managing the networking sponsorships when Roberts left ARPA. Kahn was the calm and stimulating mediator at all those meetings. Later Cerf also joined ARPA, assisting Kahn in the active development period during the rest of the seventies and into the eighties. Cerf’s patience and keen care for precision and perfection were instrumental also.
|10 – 11 Aug 74||On ferry Stockholm - Åbo|
|4 – 5 Sep 75||Linkabit Co, San Diego, California. Host: Irwin Jacobs|
|12 – 13 Nov 75||UCL, London, England. Host: Peter Kirstein|
|12 – 14 Feb 76||DCA and ARPA, Washington, DC.Host: Bob Kahn|
|29 – 30 Apr 76||BBN, Cambridge, Massachusetts. Host: David Walden|
|29 – 30 Jun 76||NDRE, Kjeller, Norway.Host: Yngvar Lundh|
|23 – 24 Sep 76||UCLA, Los Angeles, California. Host: Leonard Kleinrock|
|9 – 10 Dec 76||UCL, London, England. Host: Peter Kirstein|
|10 – 11 Mar 77||Comsat, Washington, DC. Host: Estil Hoversten|
|8 – 10 Jun 77||NDRE, Kjeller, Norway.Host: Yngvar Lundh|
|17 – 19 Aug 77||Linkabit, San Diego, California. Host: Irwin Jacobs|
|31 Oct – 2 Nov 77||BBN, Cambridge, Massachusetts. Host: Bob Bressler|
|1 – 3 Feb 78||UCLA Los Angeles, California. Host: Wesley Chu|
|3 – 5 May 78||UCL, London, England. Host: Peter Kirstein|
|31 Jul – 2 Aug 78||MIT Lincoln Lab, Lexington, Mass. Host: James Forgie|
|1 – 3 Nov 78||Linkabit, San Diego, Caliornia. Host: Estil Hoversten|
|8 – 11 May 79||BBN, Cambridge, Massachusetts|
|4 – 7 Feb 80||SRI, Menlo Park, California|
|14 – 15 May 80||MIT Cambridge, Massachusetts|
|7 – 9 Oct 80||UCL, Royal Signals and Radar Est. Malvern, England|
|28 – 30 Jan 81||ISI, Marina del Rey, California|
An active collaboration between research groups unfolded during those – about eight – years. Each group typically worked with one or more others on some special aspect. Hence, the UCLA group studied traffic and queuing, we at NDRE performed packet speech experiments via satellite, in near collaboration with MIT Lincoln Lab and so on. Daily communication was handled by the very useful new medium – electronic mail. About every third month one or more persons from each group met for a two- or three-day meeting, venue rotating between them. Results, ideas and problems were thoroughly discussed. The largest group was at BBN. They were actually implementing the experi-ments by programming the IMPs and TIPs. A very powerful experimental environment was thus created. As an example, I could conduct an experiment concerning packet transfer over satellite links, from my terminal at Kjeller, and receive detailed results there. We had an experimental satellite network between three ground stations, in Maine, England and Scandinavia. The Norwegian Telecommunications Administration contributed access to their ground station and the use of one “Spade-channel” in the Intelsat IV satellite. Various types of information transmission media were studied. Terrestrial radio nets were available in the San Francisco bay area and at Hawaii. Bob Metcalfe invented a local cable system called Ethernet. Ideas were followed by comprehensive analysis and simulations, then experimental verification. Traffic for the experiments was generated by programmed traffic generating host computers, where the intensity – packet flow – could be varied from zero through saturation of the channels as part of the experiment. And “natural” traffic between hosts in projects where resource sharing collaboration went on, was also used and provided practical cases for comparison. It took considerable effort to arrive at a system that avoided deadlock caused by packets being queued up in the nodes. Comprehensive studies went into defining suitable ways to define in formal terms the needs of varying service types. As an example, there are different needs for maximum delay and acceptable inaccuracy for speech transmission compared to those for funds transfer. Formal definition of the resulting algorithms and protocols were written and revised regularly as the development proceeded. Jon Postel wrote most of them. More and less informal documentation was issued in a series of notes called Request For Comment – RFC. They are available over the net. The protocol defining documents gradually became more stable. About 1981 they had effectively been developed into standards. The most important ones became known as Transport Communication Protocol – TCP – and Internet Protocol – IP.
The name Internet came about when it was realized that the various transmission media warranted rather different algorithms for packet handling. So each medium was treated as a separate net, and the separate nets were interconnected through gateway functions into a net of nets. IP regulates the way in which packets find their way. TCP re-establishes the original complete string of bits being transferred, into the resulting one according to the requirement of that particular service type in terms of urgency, correctness etc. Another important feature is the use of mask shaped topology of the network rather than star shaped. It means that there are alternate routes. Those protocols will route packets optimally according to the current net configuration and traffic situation. Some time in 1983 the centrally controlled Arpanet and old methods were discontinued. Since then it is all Internet. Through the 1980s a growing interest in Internet caused its expansion in the academic world, world wide. When commercial activity was permitted, from 1991 and the World Wide Web was implemented in the Internet, the net really began to grow. The Web was invented by Tim Berners-Lee at CERN in Switzerland. Around 1994 Internet began to be a recognized word in the general media. In the second half of that decade the established telecom operators accepted the highly untraditional methods of the Internet as being useful.
A great many persons and groups participated in the development of the Internet, many more than have been named here. I have only mentioned some of those that I came into contact with. I am in no doubt that those that I have named certainly deserve to be mentioned for their ingenuity and essential role in developing the powerful technology that underlies the strength, versatility, growth and practical usefulness of the Internet – of its success! As I see it, the excellent result of that development was not only the fruit of ambitions, creativity, combined dedication of many, and their meticulous investigations. It was also a good courageous investment of many years of hard work to hit a goal at the right time. That is before anybody else would have believed it.
Tønsberg, Norway October 2012. Yngvar Lundh