Dennis Jennings moved to the US in 1985 to manage the NSFnet project at the National Science Foundation. Its initial aim was to link selected research centres to supercomputers, but Dennis extended this concept into a network of networks for all academic researchers, powered by the little-known TCP/IP networking protocol. Internet services were born with NSFnet.

Based at University College Dublin, Dennis was also a leader in academic networking for Europe. He subsequently assisted the creation of campus companies, invested in Irish software firms and co-founded a venture capital business.

I spent most of my career building computing and networking infrastructure for the academic research community. I joined University College Dublin (UCD) as the first full-time Director of Computing in 1977. I brought computing and networking there from the mainframe and punched card era to the Internet / Web era with thousands of computer workstations by the time I left in 1999. I also contributed, mostly while on leave of absence from UCD, or in a part time capacity at UCD, to the early development of computer networking for research both nationally and internationally. In 2014 I was recognised by the global Internet Society as an Internet Pioneer (one of only 33 people recognised worldwide to date), and inducted into the Internet Hall of Fame for my work on the early stages of the development of the Internet.

I am a physicist by education. In 1971, while I was completing my PhD at UCD on high-energy gamma radiation from pulsars (neutron stars), I started work as a junior programmer in the UCD Computing Laboratory, advising academics and researchers on the use of computing tools. In 1972 I joined System Dynamics as a junior computer consultant. I learned a tremendous amount at System Dynamics – about people, technology, business, innovation, project management, finance, and other real world stuff – that I never learned at university.

My interest in computer networking was initially simulated by some consulting work that I did on the very early branch data collection network in Allied Irish Banks. In 1976, Michael Purser, who had worked with me in System Dynamics and had then moved to Trinity College Dublin (TCD) as a computer science lecturer, organised a conference on computer networking. The conference brought many of the data network gurus from around the world to TCD to talk about packet switching and circuit switching and datagrams, networking protocols, the ARPANET, and the early UK academic networks. I was absolutely fascinated.

So, when I was hired as the Director of the Computer Centre in UCD in 1977, one of my interests was to explore how networking might be relevant to the support of academic and research computing. My main role was, of course, to develop the UCD computing infrastructure, but in 1979 Michael Purser and I came up with the idea that we should link the mainframes in the two universities. We persuaded Dr Jim Mountjoy at the NBST to fund the project, and in due course the project team built an X.25 software switch and connected the two mainframes with a single 9.6 Kbps leased line. As far as I know, this was the first piece of mainframe-to-mainframe networking, based on non-proprietary standards, that was done in Ireland.

In 1982 I made a proposal to the Higher Education Authority (HEA) to link all the Irish universities together over leased lines to UCD. That proposal was quickly shot down, because none of the other university computer centres wanted UCD as the network hub. But, in 1983, supported by the HEA and the other universities, I made another proposal for an inter-university network to enable access to computers and to provide file transfer, data sharing and electronic mail. Funding was provided, and HEAnet was started using 9.6 Kbps leased lines. HEAnet used the UK Coloured Books protocol suite, the most advanced non-proprietary networking protocols at the time.

In the late 1970s and early 1980s, my work in developing the computing services in UCD allowed me to visit and develop contacts with many of the leading US and UK university computing centres. In 1983, Ira Fuchs, VP for Computing at the City University of New York, and one of the founders of the BITNET network, arranged an invitation for me to join the first meeting of the European Academic Research Network (EARN) at CERN in Geneva. I joined the meeting, arranged that Ireland would become a founding member of EARN, and got myself elected as the first President (Chairman). As is often the case in international fora, being English speaking and from a small country is a significant help in such leadership contests.

The EARN network, which was an extension of the US BITNET network, and largely funded by IBM Europe, was based on IBM proprietary protocols and used extremely expensive 9.6 Kbps international leased lines. Throughout the 1980s EARN was tremendously successful in providing e-mail and file transfer services to tens of thousands of ordinary academic researchers by connecting IBM mainframes and Digital Equipment VAX computers together over leased telephone lines. EARN’s early days were difficult, because the national Postal Telephone & Telegraph (PTT) providers were monopolies and they considered international data transfers on private networks between third parties to be illegal! Persuading the PTTs that the European universities could be considered to be an international academic consortium eventually finessed this third party traffic issue.

In 1983, the UCD Computer Centre won a contract to provide electronic email/conferencing connectivity for participants in the European Commission’s ESPRIT research programmes. Called EuroKom, this was the first host service to connect to Euronet, the then new Europe-wide X.25 network. EuroKom became a university company, and was eventually spun out from UCD with John Conroy as managing director.

In 1984 I was invited by Ira Fuchs and Larry Landweber, another Internet Hall of Fame inductee, to consider applying for the post of the director of CSNET. In due course I was offered that job, but I was also offered the position of Programme Director for Networking at the US National Science Foundation (NSF), which I accepted. UCD very generously offered me leave of absence, and I started work at the NSF on 2 January 1985. While I was there I was responsible for the decisions that created NSFnet, the network of networks, or internet, that eventually evolved into the Internet.

The original motivation for the NSFnet was to provide remote access to supercomputers. The insight that I had – and persuaded the NSF to agree to – was that a general purpose network for all science and engineering research would serve the supercomputer users just as well as a dedicated supercomputer access network. I proceeded then to start building a network of networks for all US academic researchers. The key decision was that all the networks and all the computers on the NSFnet would be required to run the TCP/IP protocol suite.

NSFnet model as presented by Dennis to the National Science Board at the NSF in January 1985.

NSFnet model as presented by Dennis to the National Science Board at the NSF in January 1985.

The network of networks was based on a three level model:
1. Campus networks, funded by the campuses;
2. Intermediate networks, at the NSF supercomputer centres, or regional or State networks co-funded by NSF, with one, the JvNC network at the John von Neumann supercomputer Center, running at 1.544 Mbps; and
3. A backbone network.

On 17 September 1985 – a key date in the evolution of the Internet –  the directors of the five NSF supercomputer centres agreed to host the NSFnet (interim) backbone. This would run at 56 Kbps, the same line speed as in the ARPANET.

The TCP/IP protocols were first published in 1974 by Cerf & Kahn, but they didn’t gain any real traction in wide area service networks until the 1980s. In January 1983 the ARPANET transitioned to TCP/IP. CSNET also started to use the protocols in 1983. The TCP/IP protocol suite really took hold with the NSFnet programme. By this time the term Internet was coming into use in the US to refer to the collection of local and wide area networks that connected academic desktop workstations across the US to each other and to large scale host computers.

By the time I left the NSF at the end of March 1986 I had spent or committed $17 million of US government money. But I knew that I had started something big for the US academic research community. Little did I know then just how big this Internet was to become.

European academic research networking was unfortunately way behind the US at this stage, and it took years to catch up. Here in Ireland, HEAnet decided on a private X.25 network of 9.6 Kbps leased lines (following the example of the UK), because the universities simply could not afford either the costs of, or the uncertainty of the costs of, growing the network traffic using the public switched X.25 network. HEAnet later used non-volume charged virtual circuits over the Irish public X.25 network to manage costs.

European research networking was dominated by the ISO/OSI networking standards and the use of public switched X.25 networks. The model of networking was a telematics one, designed to support low volumes of traffic from terminals to on-line databases. This was extremely heavily influenced by the conditions imposed by the European Commission on its funding for research networking, and its focus on industrial relevance, rather that on simply supporting academic research. It took forever for the research networking people involved to recognise that the public X.25 networks could not provide the bandwidth required, that the costs were prohibitive, and that the ISO/OSI protocols were inadequate for the sort of desktop workstation to desktop workstation networking (across wide area and local are networks using a variety of technologies) that was so successfully supported by TCP/IP.

Things started to change in Europe in the late 1980s when local networks of Sun workstations arrived on the campuses. These systems had TCP/IP networking pre-installed, and academics started to bypass the national networking organisations and to connect departmental LANs to remote LANs with leased lines. Soon the computer science community began to demand that the European academic networks support the Internet protocols. This led to the Ebone project, which drew up plans for an Internet backbone in Europe. I was actively involved in Ebone: I went to the start-up meeting in 1991 at Amsterdam zoo, and then served on the management committee of Ebone for a couple of years. HEAnet moved to 64 Kbps leased lines running TCP/IP in 1992.

The development of the World Wide Web in CERN was the next major leap forward. UCD Computing Services ran an early trial of the web protocols, but it wasn’t until 1994 that the use of the web accelerated. Academic researchers were very much at the leading edge of Internet development and use. The Mosaic web browser was invented at one of the original supercomputer centres that had connected to the NSFnet interim backbone in 1986.

In 1991, I contacted Jon Postel at the Information Sciences Institute at the University of Southern California, who had been administering the domain name system for the Internet. He and I agreed that UCD Computing Services should take over the administration of the .ie Irish country code top-level domain name (ccTLD), and that was the start of the IE Domain Registry (IEDR). All the IEDR policies are rooted in the decisions that I made in the 1990s. These decisions were driven by the idea that the .ie ccTLD is a valuable asset for the economy and that names in .ie should be restricted to entities that have a real and substantial connection to Ireland rather than made available to anyone. While it took a long time to get the price of an .ie domain name down to reasonable levels and to get the systems right, I believe that that policy has served Ireland well. It is one of the few top-level domains in the world where you can be pretty sure that there is a real and genuine entity behind a name.

By 1995 the idea was well established that a host computer could be accessible globally and could offer services to people anywhere. I realised the commercial potential when a group of UCD computer science researchers walked into my office and asked for assistance to establish a web-based training company. Their prototype application had taken first place in the client/server education and government category of the Apple Enterprise Awards contest in New York. Together we set up WBT Systems as a campus company in 1995. I became its chairman and I funded the company on the back of my credit card for the first eighteen months until Paul Kenny joined as MD and raised its first round of venture capital from Delta Partners.

By way of contrast, I was invited at around that time to speak to the Irish Business and Employers Confederation’s telecommunications group. I told them that all computers were going to be connected through the Internet, and that all the other networks were going to disappear, and that all applications were going to be applications on the network, using the Internet protocols and the World Wide Web. The telecommunications and business people at the meeting looked at me as if I had at least three heads. They had no idea what I was talking about.

In addition to all my academic support responsibilities, which between UCD Computing Services full time job and my part-time computer networking activities had me working very long weeks, I became increasingly interested in business entrepreneurial activity and technology start-ups. In addition to WBT Systems, I assisted a small number of companies spun out from UCD. More significantly, back in 1990 I was the first investor in Jim Mountjoy’s new company, Euristix. Jim led this company to become a significant employer of software development people in Ireland, and to international success. In 1999, the company was sold, and with the proceeds from the sale of my by then very small shareholding in the company, I was able to begin to explore a new career, and I left UCD.

Last edit: October 2015

© Dennis Jennings 2015