Thursday morning, I sat on the track at the Utrecht station waiting for my train to Bonn. If you’re a few minutes early, its fun to go up to the tracks and watch the other trains leave. The 9:23 local was leaving on my track and I watched the clock turn to 9:23. Exactly two seconds later, the train was on its way.

One of the questions that always puzzles me is what conditions lead to a successful infrastructure. The European networking infrastructure was certainly not as developed as it could be. Likewise, the telephone infrastructure had always been expensive and not always on the forefront of technological change.

Yet, the European passenger train system, developed in a highly fragmented period of highly sovereign governments, was a miracle. The trains go to every town of consequence and many of none. Even more remote places are connected to the transportation web by a series of bus systems, coordinated with the train schedules, of course.

Staring at the track, hoping for a flash of insight, I realized that my train was about to leave. A few hours later I arrived at the Hauptbahnhof in Bonn and pulled out my itinerary of travel reservations. To my horror, I discovered that I was booked in a suburb of Bonn, Königswinter.

My instructions to my travel agent had been simple. “No Holiday Inns, no suburbs.” Bonn, like Prague, is one of those cities that never has enough hotels and I was lucky I wasn’t as far away as Cologne, the provisional capital.

I wandered around the train station for a while trying to figure out how to get to Königswinter. I spied a sign for the tram system with the magic word on it and caught the first one leaving. The tram was labeled Bad Honef and I crossed my fingers that the track didn’t serve multiple locations.

When I got on, I realized that I didn’t have a ticket. All over, colorful signs in multiple languages indicated that tickets must be immediately validated or I would face a DM 60 (U.S. $36) fine. Having no idea how to get a ticket, or where I was going, I alternated between peering out the window looking for some appropriate sign and looking nervously around for some sign of the ticket police.

Königswinter turned out to be a resort on the Rhine river. My hotel had a view of four castles up in the hills, was right across the town square from the church belltower, and my room was right on the river, looking out at barges and ferries, the tram, a boardwalk, and people out strolling in the cold air.

I went down to the sun room to work. Old men, couples, and mothers with children came in periodically to get a cup of coffee, a beer, or some ice cream. I ordered a cup of coffee, with the waitress checking to make sure I just wanted one cup.

The next morning, at breakfast, the waitress wanted to know if I wanted my coffee in a “can or a cup?” Outside it was still dark and I could make out the fog banks in the hills. Suddenly, the fog started to glow and I could see the black silhouettes of the castles. Then it was daylight and time to try to make my way into Bonn.

I took the tram back into the main train station and went in search of an information booth. I waved a business card in their face and they sent me to track 2. As I was trying to decipher the map of the extensive tram system, one pulled in. I suddenly realized that I had been sent to the wrong track, narrowly missing getting on one going the wrong way.

Armed with a valid ticket, I arrived at the stop in front of the monolithic police station headquarters, walked a few blocks over to the offices housing GMD, and found Klaus Birkenbihl.

Klaus is the manager of the Netzzentrum fuer die Wissenschaft, the science networks center in the Institut für Anwendungsorientierte Software- und Systemtechnik at the Gesellschaft fur Mathematik und Datenverarbeitung mit beschrankter Haftung. In the interests of saving space, it is hoped the reader will excuse the informality of referring to the organization as simply GMD.

GMD is a government-owned research institute specializing in mathematics and computer science. Like INRIA, the French institute, the institute gets 60 to 70 percent of its funds from the government, raising the rest through contracts with industry. GMD has 1,200 staff, of which 700 are scientists.

I had come to see Klaus to find out about one more network, the IBM-sponsored EASInet. I had always assumed that BITNET, and its European cousin, EARN, were IBM’s contribution to networks. I heard periodically about EASInet, but did not know anything about the network nor why IBM should start yet another network. EARN had started in 1984, originally with six countries but, after Dennis Jennings heard about it, six countries plus Ireland. The IBM model was to fund the program for three years, after which it would become self-sustaining. EARN added many countries and did, in fact, become self-sustaining.

Not only did EARN become independent, but you might even say the board got a touch hostile, planning a transition to OSI and cozying up to DEC. The OSI focus was actually a political necessity, part of a compromise with the PTTs who agreed to loosen some of their restrictions on clients sharing leased lines in return for a promise by EARN to migrate to OSI as soon as the protocols became technically and financially viable. This all happened in 1984 and the OSI migration took a bit longer than anyone expected. By the time I got there, most EARN sites were migrating to the BITNET 2 protocols, more commonly known as TCP/IP.

The upshot of the OSI (and later TCP) focus was that it was quite obvious that EARN was no longer an IBM animal. Creating a network that was self-sustaining after three years was a great public service, but didn’t really serve the other goals of maintaining the IBM presence in Europe and, in the long run, selling more iron.

IBM developed a new proposal as a way of providing an incentive to the big mainframe market in the research community known as the European Academic Supercomputer Initiative (EASI). Under the program, IBM plunked down large amounts of heavy metal in selected research institutions and paid for the cost of 64 kbps links between participants, the collection of those links forming a fairly hefty donation of bandwidth. In addition, IBM paid for the cost of a T1 line from CERN to NSFNET and hired GMD to manage the network.

EASInet consisted of 18 sites by the beginning of 1992, with locations from Spain and Italy in the south, to Stockholm, Amsterdam, and Hamburg further north. Most of these sites were not exclusively “EASInet” sites, but were on a variety of networks. CERN, for example, an EASInet site was sometimes referred to as the Center for European Research Networking.

In a few cases, the EASInet link was in fact a dedicated 64 kbps line between two sites. In most cases, however, the IBM money had been thrown together with other sources. For example, a 256 kbps line between Amsterdam and CERN was split between HEPnet, EUnet, and SURFnet. The line sharing arrangements that EASInet started were an important early example of the model later used in EBONE.

To handle multiple protocols, many of the EASInet lines used NET’s IDNX time division multiplexers, with most of the bandwidth given over to TCP/IP and SNA and other, smaller, circuits applied to X.25 and DECnet. EASInet was really bandwidth. In some cases, GMD actually took care of the lines, in others the job was split with other organizations. Some of the data flows were point-to-point sessions, others were part of broader networks such as HEPnet or the European Internet.

When IBM had agreed to cooperate with EBONE, what they meant was that the T1 line to NSFNET would become a pathway to the U.S., and thus to Asia. Even though IBM had not signed the EBONE document, the link would be available to transit traffic.

Helping coordinate EASInet was only one of GMD’s projects. The site also acted as a key gateway between Germany and the rest of the Internet. Germany had long been one of the most ardent OSI advocates. In 1984, the ministry of research and technology started up a membership organization, the Deutsches Forschungsnetz (DFN), the German research network intended to give the German scientific community the full benefits of OSI.

GMD, part of the federal research establishment, ended up with a sort of schizophrenic role. On the one hand, they took part in the fervent OSI cult built around the leadership of their funding agency, helping DFN on conceptual work and software development.

On the other hand, GMD had work to do and actively helped build EARN, running the German national node since 1987. At the same time, GMD started building a private X.25 network to link German national research laboratories together.

This private network, AGFnet, was not OSI (in fact it was SNA), but at least it contained X.25, the “pathway to OSI,” to make it politically palatable to the bureaucracy. What AGFnet did do was prod DFN into action, which resulted in a national X.25 network called Wissenschaftsnetz (WIN or “science network”).

DFN had a notable success in building WIN: they made it affordable. With the fairly strong backing of the federal government, DFN was able to convince the German PTT to give it preferential rates and, most importantly, to avoid imposing volume-based charges. The result was a rate of DM 5,000 per month for a 64 kbps line. By 1991, WIN had spread throughout Germany and much of the research community was on the network, including commercial researchers such as Daimler Benz.

After what Klaus wryly termed “a period of difficulty,” DFN accepted the premise of multiprotocol networks. In fact, with users paying for their own usage, it turned out that the vast majority of sites chose to use TCP/IP or EARN protocols.

There was still some OSI work on the network, particularly since the federal government was willing to finance such research. X.400 usage, in particular, had significant usage and was growing quickly. GMD operated application gateways to the Internet and EARN worlds to keep connectivity.

Hanging in Klaus' office, high up near the ceiling, was an abacus with a little mouse perched on top. This odd little arrangement was the Birkenbihl massively parallel supercomputer, given to Klaus by his daughter. “Look,” he said, pointing to the blue and red beads, “it even has a color graphics display.”

I went back to Königswinter, hoping to warm up with a nice, hearty German meal, but it was the off-season and all the interesting places were boarded up. A half-dozen Greek restaurants appeared to be Gyros-only joints, with nary a drop of taramasalata in sight, so I settled on an Italian restaurant. There I was served limp, overcooked noodles in a heavy, congealed green sauce.

“That was the worst pesto I’ve ever had,” I told the waiter as I paid up.

“Thank you very much,” he said with a proud smile.

I finally found a sausage shop open and filled up on a couple mugs of Konig-Pilsner and a Riesenbackwurst, a greasy sausage served on a slab of cardboard with a puddle of mustard. Much better than the lime-green pesto.

The next morning, checking out of the hotel, I watched the clerk punch my room number into the computer and we stood staring at each other with forced smiles while the dot matrix printer chugged away. The clerk retrieved the printout, then walked over to an ancient cash register and punched in the room charges, one by one. He then picked up a thick sheaf of bar bills and added them in, one by one.

He took that printout over to a calculator, where he added in my minibar charges. By then, I was not in the least surprised when he took that slip over to his desk and figured out the tax, adding it in by hand.

“Not very modern,” he said with a shrug.

I had to agree.