Introduction - An overview of this report
History - How it all started
TCP/IP and the Internet - Why it is significant
Conclusions - summary of the main points and general conclusions
References - The websites and books referred to in this report
My report plan
Notes - Links within the text that will take you to these notes are RED
In the 1970's satellite networks such as SATNET and radio networks, such as ALOAHNET were being developed. It was clear that some way had to be found to link these different technologies to each other, as well as to the fledgling terrestial networks. The International Network Working Group, which amongst others included the American Internetting Project, the UK's group at the National Physical Laboratory and the French CYCLADES project, realised that a means of linking these networks had to be found. This meant that a new protocol had to be invented. Bob Kahn, formerly of BBN, was working for Larry Roberts at DARPA by 1972. Vint Cerf was working at UCLA and was on the Network Working Group. These two were destined to have the most profound effect on the development of the Internet. Their collaboration was to produce TCP/IP - (Transmission Control Protocol/Internet Protocol. Click here for an explanation by Howard Gilbert of Yale University.) Without it, networks would not be able to communicate with each other.
[Source: Courtesy of TECHNOS Quarterly]
Vint Cerf was a UCLA graduate who had attended Stanford on a scholarship from his father's company. He graduated in 1965 and took a job with IBM for a short while, before joining UCLA's computer science department. Whilst at UCLA, Cerf was very heavily involved in writing the software required for the host to host communications. He was also a member of the Network Working Group, which developed the first fully working network protocol known as the Network Control Protocol or NCP.
[Source: Courtesy of CNRI]
Bob Kahn had been a professor of engineering on leave from MIT and working at BBN. He had a special interest in networking and was encouraged to work on it by BBN. It was Kahn who had urged Larry Roberts (Director of ARPA) to conduct the ARPA networking experiment on a large scale. Kahn, himself, actually wrote the host - to - IMP specification, which enabled the UCLA team (which included Vint Cerf) to complete their part of the project.
Early on in the design of the ARPANET Kahn had been concerned about system 'lock-up' and was able to prove that his fears were well founded during January 1970, when he and Dave Walden tested this out in Los Angeles. The crucial thing about how the net was to work involved packet switching and the reassembly of these packets appeared to be the problem. While Kahn was grappling with this problem he worked closely with Cerf for the first time.
Kahn left BBN and joined DARPA (The D for Defense had just been added) in 1972. It was at this time that Kahn started thinking about linking different networks together. Just prior to this Kahn had been involved in the ICCC demonstration of the ARPANET with Vint Cerf. Cerf was in charge of the International Networking Group (INWG) of which ARPA's Internetting Project was a member. Other Packet Switching projects included Donald Davies's National Physical Laboratory in the UK and Lois Pouzin's CYCLADES project in France. All were enthusiastic about developing network-interconnection technology. The INWG "began pursuing what they called a 'Concatenated Network' or CATENET for short, a transparent interconnection of networks of disparate technologies and speeds" (Where wizards stay up late)
Early in 1973 Kahn and Cerf began talking seriously about the problems involved in network interconnectivity. They decided that they needed 'gateway' computers which would route messages from one system to another, using software that made them appear to be hosts to the respective networks. The Network Control Protocol, used by the ARPANET, would have to be replaced by a more independent system. "The challenge for the International Network Working Group was to devise protocols that could cope with autonomous networks operating under their own rules, while still establishing standards that would allow hosts on the different networks to talk to each other." (Where wizards stay up late)
Cerf and Kahn spent most of 1973 working on the problem. They finally presented a paper in September to the INWG at a meeting in Sussex, England. Much discussion took place with members of Davies's and Pouzins teams.
The Cerf-Kahn paper was finally published in May 1974. Under these proposals packets would be sent within 'datagrams' much like a letter within an envelope, with the gateways only reading the envelopes, so that the packets would be 'taken out' and read by the receiving hosts. The system would be regarded as unreliable, in the same way as CYCLADES was, so if a packet didn't arrive or was unintelligible, and no acknowledgement was received, a 'twin' was transmitted. This meant a focus on end-to-end reliability rather than reliability from the network itself. Work continued on polishing it into a workable protocol, until 1975 when the TCP specification was sent for implementation to BBN, Cerf's Stanford group and University College, London.
During 1976 Cerf finally joined DARPA as a program manager, responsible for packet radio, packet satellite and research on the ARPA Internet. He also continued work on the TCP specification. In October 1977 Kahn and Cerf were involved in a milestone demonstration when they linked a three network system of packet radio, SATNET and ARPANET. This involved sending packets over 94,000 miles and there were no transmission errors. In early 1978 Vint Cerf, Jon Postel and Danny Cohen discussed the idea of removing the part of TCP that dealt with routing packets and have this done by a new protocol known as Internet Protocol. TCP would then be free to concentrate on breaking up messages into datagrams, the reassembly process, error detection, putting packets into the right order and resending any missing packets. This idea came from the Xerox team at Palo Alto who had done something similar, on a smaller scale. Thus TCP/IP was finally born in 1978.
By the Spring of 1983 TCP/IP was fully implemented on the ARPANET, with NCP no longer in use. Many other networks were also using TCP/IP. Now they could communicate with each other. However, TCP/IP had a rival - OSI (Open Systems Interconnection), developed by the International Standards Organisation. Most governments wanted to use it as the world standard communications protocol. The problem with OSI was that it was just a specification and not a tried and tested system as TCP/IP now was. TCP/IP was, after all, an 'open' design refined by large numbers of computer professionals and academics. It had also been adopted for use on Ethernet systems (hardwired local area network systems), such as would be found on a university campus. There was almost an underground movement, within the computing community, that quietly championed the cause. Another important developement was the adoption of TCP/IP by the inventors of UNIX, a very popular operating system. Unix was used by SUN Microsystems workstations. When SUN provided the networking software free with their machines networking took off. The march was relentless and by 1989 virtually everyone was using TCP/IP.
By 1994 the Internet had become a huge concern, partly due to the explosion of the World Wide Web (invented by Tim Berners-Lee) and consequent use of HTTP. It had started to find its way into everyday use for many households and businesses. However, the 'glue' that held all this together was TCP/IP. Without it, none of the many networks that made up the Internet would be able to communicate with each other. TCP/IP also meant that these communications were handled in a very reliable fashion. It's 'end to end' reliability meant that any problems with any particular network would not cause any difficulties with the message itself.
I believe that it is very significant that TCP/IP is an 'open' system. It very much characterises how the Internet itself works - as a kind of democracy. It is also significant, from a technical point of view, that it is able to link many disparate networks in a simple and very reliable fashion. Some of these are Local Area Networks and others are huge transcontinental systems, but they all have one thing in common - TCP/IP. Therefore, no 'translation' is necessary. It is possible for anyone to communicate with anyone else, with no barriers. Information can be passed freely and easily from one part of the globe to another.
The companies, organisations and people mentioned in this report can be found at:
Hafner K. and Lyon M. (1996) Where wizards stay up late. Simon and Schuster
TCP/IP - Click here for a very detailed description in Adobe Acrobat format (.pdf) [Return].
ARPA - Advanced Research Projects Agency [Return].
ARPANET - The first computer network [Return]
UCLA - University of California at Los Angeles [Return]
MIT- Massachusetts Institute of Technology [Return]
BBN- Bolt, Beranek and Newman - The company awarded the contract by ARPA to create the ARPANET. [Return ]
IMP- Interface Message Processor - A computer used to connect a host computer to the ARPANET. [Return]
HTTP- Hyper Text Transfer Protocol [Return]
PACKET SWITCHING- This very simplified diagram shows the structure of a packet:
Source: Open University
[Return to text]
Click here to return to top of page
Return to home page