I joined the UCD Computer Centre in January 1979 following three years of postgraduate research into stability of dynamical systems. At that time the centre had about two dozen staff. Dennis Jennings had become its director and was actively extending computer access across the campus. Computing services had previously been concentrated in just one or two privileged locations. The new policy brought a shift from batch processing to timesharing. I worked at first in user support, advising people on how they could use the systems, giving courses on application software, and developing new services. After a couple of years, I moved to the systems programming team.

At the start of the 1980s, before HEAnet began to take shape, our only external connections were over a couple of leased lines: one between Belfield and Merrion Street and another to Trinity.

During my early years at the Computer Centre I worked on a skunkworks project that made it possible to print DECSYSTEM-20 output on IBM printers. It involved making components that were already there into an integrated system. That type of problem solving became a characteristic approach for me.

When Ahmed Patel and I did a study on the available options for campus networking in UCD, we looked at how network technology was shaping up and what actually worked. That led us to recommend X.25 at the lower level. I found myself on the HEANet technical committee planning deployment of X.25 boxes for the first phase of HEAnet – at that stage a collaborative effort between the five University Colleges and the two National Institutes of Higher Education.

X.25, however, was a mess. The national monopoly telephone operator issued X.25 numbers and used a numbering scheme which reflected the capacity of the line. If you upgraded your speed you had to change your address because 1200 bps or 4800 bps or 9600 bps was encoded into the address. It was just crazy.

Back in the early 1980s it was somehow recognized that communications in the academic community were different from communications elsewhere. Things that were forbidden in the commercial world could be tolerated there. This may have been because academics were doing networking research as well as doing research communications. The research might result in new opportunities for the phone companies downstream. This was surely a form of doublethink. And it meant that commercial data networking lagged behind academic networking – as well as behind what was happening on in-house corporate networks.

In 1984 IBM decided to put some money into something useful for Europe. It supported the launch of the European Academic and Research Network (EARN), which was based on proprietary IBM protocols and linked to the US academic and research network BITNET. Dennis Jennings was to serve two terms as president of EARN. The network’s initial sites in Ireland were UCD and University College Cork, connected over a leased line connection. Another leased line connected the ‘National EARN Node’ at UCD upstream via the UK National Node to the rest of EARN. I became the country co-ordinator for EARN.

Olivier Martin has written in his book ‘The “Hidden” Prehistory of European Research Networking’ that EARN was the predecessor of the internet in Europe. This was so largely because the protocol it used, although proprietary to IBM, was in practice open and widely emulated by other suppliers. The resulting ‘vendor neutrality’ allowed EARN to create a very open community for academics.

We soon had e-mail that went beyond the campus and were able to collaborate with partners in other countries without the high costs of telephone calls or the delays of the postal system. Once you knew a destination node name, you could tag a file with certain control information for delivery to a printer or a person in Berlin or Stockholm or CERN or Montreal or Stanford. EUnet had similar capabilities but, being Unix-based, seemed more esoteric.

As EARN grew it came under pressure from Brussels, especially through the Eureka project known as Cosine, to develop an OSI-based backbone. We were also aware that the service needed sustainable funding, as IBM would end its support at some stage. In October 1987 Digital announced that it would support the migration of EARN to OSI. Northern Telecom agreed to provide a number of X.25 switches. EUnet came into the project as a collaborator. So did Nordunet, which ran a multi-protocol network across Scandinavia and Finland. RARE, the association of European research networks, was not directly involved in the project, but was significant because its Working Group 4 (whose Secretary was Michael Nowlan) was guiding the Cosine implementation.

Digital provided manpower for the OSI project, a national node (G-BOX) in each of the dozen or so EARN countries taking part in the project, and rented office space in the Netherlands. EARN agreed to contribute a techie or two. One of those was me. I moved, with my wife and two sons, to Amsterdam in summer 1989.

I went into the project without knowing how bad the politics were. Digital was mainly interested in displacing IBM. The telecommunications monopoly was also a big problem. The national ‘administrations’ (as the PTT operators were formally called) had a stranglehold on address allocation and international connectivity, according to the OSI model as standardized in International Telecommunications Union recommendations. In addition, Cosine created the International X.25 Infrastructure (IXI) backbone, which the national research networks were expected to use. It all added up to an impossible set of circumstances. On a personal basis, though, the project gave me loads of exposure and I got to know a lot of very interesting people.

The EARN OSI backbone (in red) and its connections to the national EARN networks. Directly-connected OSI nodes are shown in green; those connected over IXI, in blue.  Nodes shown in orange are ones for which DEC had made commitments in advance of the decision to close down the project.

The EARN OSI backbone (in red) and its connections to the national EARN networks. Directly-connected OSI nodes are shown in green; those connected over IXI, in blue.  Nodes shown in orange are ones for which DEC had made commitments in advance of the decision to close down the project.

The EARN OSI project achievements were not to last. We did indeed put together an X.25 backbone and the software components which allowed Digital’s VAXes and IBM’s machines to connect to and communicate over this backbone. In summer 1990 EARN decided to wind down its own X.25 backbone network and to transfer the service to IXI.  The G-BOXes remained in place, each in the charge of its national EARN community.

I returned to UCD in July 1991 to work on a series of development projects, the first of which was HEAnet-2, a multi-protocol network based on X.25, over which IP and IPX were to be encapsulated.

UCD now had two IXI connections – one serving the HEAnet sites and a second for EuroKom. These also gave external Internet connectivity due to a characteristically simple and subversive initiative from RIPE chair Rob Blokzijl; he offered any IXI participant an X.25-encapsulated IP connection to the internet via the IXI access point he managed at the NIKHEF in Amsterdam.

The internet’s impact on UCD was gradual and tapered. In some ways, it was ‘business as usual’ for network services; because of EARN, a significant percentage of the academic staff was already well accustomed to online chat sessions, transferring files and exchanging e-mails. When internet access came along, the scope of this service simply became wider. And when the web arrived, it also appeared at first to be just an extension of earlier things. HTML with its tags and content seemed a small step beyond the mark-up conventions used by the document-processing applications of the day.

The evolution of my e-mail address over the lifetime of the EARN OSI project, superimposed on my EARN business card.  The first of these addresses is built from the EARN node name

The evolution of my e-mail address over the lifetime of the EARN OSI project, superimposed on my EARN business card.  The first of these addresses is built from the EARN node name “HASEOC” and the pseudo-TLD “BITNET”.  The second is based on the domain name obtained from the .NL registry for the EARN OSI Centre.

E-mail addresses were originally in the format user@node. In the days of EARN and BITNET there were probably no more than 3,000 or 4,000 nodes. So naming was easy to manage. At a certain stage, the idea emerged that node names should be placed in a hierarchy and we found ourselves with addresses such as user@ucd.ie. And the community of people who wanted e-mail was growing. We had to synchronize the mail tables that were used in EARN with the ones that were used in HEANet, which had started with a naming scheme from the JANET network in the UK. We also had to take account of the mail translation tables in BITNET. In short, we built up our expertise in mangling addresses.

The domain name system (DNS) predated the ubiquity of TCP/IP. Back in 1986, while Dennis was in the US, he had arranged for UCD to become the effective target of a wildcard mail exchanger (MX) record for Ireland. This meant that, in the absence of a real domain registry, we ended up running a national mail gateway.

While I was in Amsterdam Mike Norris began acting as the domain registrar. That meant that he got regular requests to set up new names, mainly from Michael Nowlan in IEunet. All these requests were passed on to Harvard where the master copy of the .IE zone was hosted.

Before I came back to Ireland, Michael Nowlan had begun to press for setting up the master name server for .IE on the ‘right’ side of the Atlantic. Piet Beertema, who managed the .NL domain, convinced me to abandon an initial plan to have the zone hosted at CWI in Amsterdam and so, towards the end of August 1991, we moved it from Harvard to UCD.

This chart, which was produced in late 1997, shows the growth in the number of .IE domains. The white strip represents the domains registered with MX records. These dominated in 1995, but had all but disappeared by February 1997.

This chart, which was produced in late 1997, shows the growth in the number of .IE domains. The white strip represents the domains registered with MX records. These dominated in 1995, but had all but disappeared by February 1997.

UCD was now actually operating the IE Domain Registry, as well having administrative responsibility.

As e-mail was then the main application, and IP connectivity was still rather exceptional, the IE registry had to hold MX records for most domains registered – identifying the respective mail servers. Only a few domains were delegated to their own name servers and so had name server records. Delegation gradually became usual, but ‘MX-only’ domains persisted for many years.

For about five years, administration of .IE remained one of the tasks performed by development or operations teams within UCD Computing Services, with me as the responsible manager.

We didn’t write down any rules until 1995. We just followed a set of practices that we thought represented consensus views on how networking resources should be made available to the target community. Each holder was allowed to register just one domain, because that seemed like the responsible way to manage the resources. This was a widely held perception at the time, not just in Ireland. The IEDR launched a consultation process to review the policy in 1998. It was also customary not to permit names that had just one or two characters – with the single exception of the University of Limerick, which had suddenly replaced ‘nihel.ie’ with ‘ul.ie’ as its mail domain in 1989.

Our rules, when I recognized that they needed to be written down, were essentially a translation from the Dutch of rules that Piet Beertema had been using for .NL. A couple of years later Bulgaria borrowed the Irish rules in a similar way, changing all the .IE references to .BG. They even left in the acknowledgements that I had given to Piet and to Michael Nowlan!

The IEDR started as just a small part of my job and that of a few other systems and operations group members (mainly, after Mike Norris’s departure to HEA, Rosemarie Power and Arthur Green), but the work got bigger and bigger. In 1996 or so, the demand had grown beyond what this group could handle along with running the campus network and we began to hire into a dedicated registry team, with Pat Kane as the first recruit. Niall Murphy followed in late 1997. He had been one of the group who started the UCD NetSoc, a student society which ran an e-mail service for UCD students before the college began to do so.

I continued to manage the IE Domain Registry until late in 1999. Shortly afterwards, UCD spun out the organisation and the new IEDR Ltd moved off campus. I remained with the College’s IT Service until 2014, working in network management and engaging in wider internet activities through RIPE and AFNIC (the French counterpart of the IEDR).

Niall O'Reilly 2016Last edit: February 2016

© Niall O’Reilly 2016