Utrecht

Wednesday morning, I brought my bags down to the train station and joined a stream of Dutch commuters. It is not unusual for people in Holland to live at one end of the country and work at the other. I was going half-way across the country to Utrecht, a half-hour away.

The train ran parallel to one of the main Amsterdam canals, lined with houseboats and barges, then past warehouses and windmills. Passing the suburb of Amstel and the sprawling IBM complex, we started picking up speed. Suddenly, we were in the countryside, an elaborate system of canals crisscrossing the fields and forming a series of little islands, connected by bridges and sluices.

It was market day in Utrecht and, since I didn’t have a hotel reservation, I headed right into the middle of the town square to look around for a place to stay. Spotting one, I avoided the temptation to stop at stands full of cheeses, herring, and baked goods and got a room.

With a few hours to kill, I brought my laptop into a little cafe by a canal, one of a half-dozen places crowded with tables of high-school and college kids sipping coffees and beers. At 2 P.M., I gathered my things and went into the shopping center complex surrounding the train station to find Kees Neggers, co-managing director of SURFnet, the Dutch academic and research network.

After seeing so many networks like EUnet in Amsterdam and AARNet in Australia run with a couple of guys and a Sun, it was certainly a change to see SURFnet occupying a whole floor of plush offices. When I asked Kees how many people worked there, he said only 20, but many of them were responsible for supervising subcontractors.

SURFnet is a company that is owned 49 percent by the Dutch PTT and 51 percent by the SURF Foundation, a group started initially by the universities, but which has grown to include research institutes, the research arms of corporations, and other institutions of higher education such as teaching hospitals and vocational schools.

To explain SURFnet, SURF, and the PTT, Kees went to the whiteboard and carefully drew the organizational structure, drawing my attention to the legal status of all the parties and pointing out the differences in Dutch law between the limited liability partnership, the foundation, and the association.

As Kees continued his chalk talk, it occurred to me that he had not said one word about the network itself. I knew that SURFnet was born as a project of SURF in 1986, had become a real company in 1989, and I even knew about things like the fact that it had a capitalization of 3 million guilders (U.S. $1.6 million), and expenditures of NG 14 million per year, of which NG 10 million were cost recovery services from users and 4 million was development paid for by the government. What I didn’t know was what the network did.

Well, it turned out that SURFnet started by using the X.25 data service from the PTT. After some long thinking and a big RFP, Northern Telecom got the contract to install and manage a private X.25 network based on 64 kbps links. This net was based on three major hubs, with each leg of the triangle having two or three 64 kbps links. Another 16 sites were equipped with switches, each of these having at least two lines into the network. Altogether, 413 sites had connections.

The network appeared to be your basic FOO over X.25 net. Most of the services had SURF* names. SURFmail, for example, was described in the annual report as “using the RFC-822 addressing mode as an intermediate step towards the international X.400 standard.”

Recently, the network had begun the process towards a 2 Mbps backbone. To quell any political problems, five sites were given 2 Mbps links and a “pilot” was run. Some lines ran straight IP, others ran straight X.25, and some ran a multiplexed mess of the two. Needless to say, the results of the pilot showed that running IP on a line was more efficient than running IP on top of X.25. However, with X.25 identified in many European countries as the “pathway to OSI” it was important to conduct the pilot.

All this technical information I learned in a few minutes from Kees and by leafing through the annual report. Very quickly, though, we were back on the subjects of organizational charts, examining how his network management staff served as contract supervisors to monitor the subcontractors that were doing network management.

We moved on from SURFnet to pan-European politics, an area in which Kees is an extremely active player. His first foray was as the EARN representative for his country. When RARE was formed, SURF was named as the Dutch national member, and Kees was named SURF’s representation to the RARE Council of Administration (CoA).

RARE, the Réseaux Associés pour la Recherche Européenne, was started in 1986 to promote the use of networks based on open systems solutions, read “OSI.” This was to be sharply contrasted to EARN, IBM’s proprietary protocol, association, and network, foisted upon unknowing European countries as part of an alleged plot to dominate the market. Well, maybe not quite that sharply contrasted, but it would not do to underestimate the fury of this particular religious war.

At that time, only the U.K. really had a national network. Many people assumed that OSI would be the basis for networking and that the PTTs would provide the infrastructure on which OSI would run. That infrastructure would be a commercial offering, universally available, and available soon.

The focus on the PTT as the service provider led to the assumption of a single national network, and thus RARE is made up of one member per country. There are also various associate members, international members, and liaison organizations, but the basic governance of the organization is by the Council of Administration which has one seat per country.

There is another element of RARE that is worth noting. It gets money from two sources: contributions from member countries and from the European Commission. The European Commission component is strong enough that they wield considerable clout on RARE decisions. Most of the linkage between RARE and the Commission has been through the COSINE project. (Every acronym must be pronounceable and preferably have a double meaning of some sort, however inane. In the case of COSINE, the expansion yields Cooperation for OSI Networking in Europe.)

COSINE was started in 1986 by another group called Eureka, a research fund subscribed to by a group of countries toughly equivalent to the RARE membership or the European Free Trade Association. The COSINE project of Eureka was handed over to the European Commission to act as project coordinator and they, in turn, picked RARE as the secretariat.

What all this meant was that Eureka kicked in some money, the Commission kicked in some more, and a fraction of that money flowed downhill to RARE. Eventually, so the theory went, money would flow down even further into a series of market-pull activities that would prepare and stimulate OSI in Europe.

Much of the research funded by the Commission was in the area of “pre-standardization research.” Pre-standardization was the process of preparing a bunch of documents that are then used as input to the standards process. In the case of COSINE, this meant establishing the subset of OSI that would best fit the special needs of Europe.

The functional profiles of OSI were prepared by a series of working groups. Every country got a member. The chair could invite four “experts,” sort of a standards patronage position. The EC paid for the travel expenses of each delegate, kicked in for the cost of donuts at the meetings, and took care of the RARE secretariat.

Paying people to go to meetings in fine places has the sure result of generating a large demand for meetings. The specification process dragged on as people tried to define which portions of OSI would be used in Europe. The assumption was that when those profiles were developed, the lure of the common European market would be enough to attract vendors like flies to a water buffalo.

Vendors and researchers took one of two approaches. A few went ahead and just implemented the portions of OSI that made sense in a product. ISODE was a perfect example of this. Most just went ahead and sold TCP/IP products.

After a couple of years of this process, a set of 10 very thick, very blue volumes of COSINE specifications were issued. These 10 blue books were summarized in a red book, which in turn was summarized in an orange executive overview. I suspect quite a few corporations took that orange distillation and concentrated it yet further until senior management was hearing summaries such as “this stuff is great, trust us.”

Once the specifications were complete, RARE developed a COSINE implementation plan. In January 1990, this implementation plan was signed as a contract between the Commission and RARE. The RARE COSINE Project Management Unit (PMU) became a reality.

COSINE was designed as a three-year plan with a budget of ECU 30 million (U.S. $36.9 million). Most of that money, over ECU 12 million, went into IXI, the International X.25 Infrastructure (IXI). A large chunk, ECU 6 million, went into the PMU.

IXI is an international X.25 network for researchers. Countries like Greece, Yugoslavia, Ireland, and Portugal all use IXI as their main path into the European mainland. Even the U.K., for some strange reason, links itself into the continent with IXI.

The network is simply a star configuration of a set of 64 kbps lines that link national X.25 networks. Typically, a private, virtual X.25 network is built on top of the PTT’s own offering. There are actually two stars, one in Amsterdam and one in Bern. Two 64 kbps links connect the stars, and each country gets one 64 kbps link into one of the hubs.

Except for the U.K., that is. The U.K. is big enough that it really needs two 64 kbps links into Amsterdam. Of course, that means that you’re pumping 128 kbps of data into a 64 kbps network, but that’s another story.

In a classic example of how networks will grab whatever resources they can, much of the IXI traffic consists of setting up a virtual circuit to NIKHEF, which operates a gateway from IXI into the Internet. While running TCP/IP on top of X.25 is not the most efficient implementation in the world, when that’s your only choice, it works fine.

The theory behind IXI, and all COSINE projects, was to stimulate the market with the initial implementation. At the end of the three-year trial, some commercial entity would see the inherent profitability of the enterprise and step forward boldly with an aggressive bid to take the project over as a going concern.

At the end of 1992, IXI would terminate, and there had not yet been a clamor for rights to run the service commercially. Unless, of course, there might be some Commission money in it. This had occurred to a few people, and RARE was about to start lobbying for an Operational Unit.

Meanwhile, a second force reared its ugly head: people needed networks to do their jobs. Most of those national networks had started carrying large amounts of TCP/IP traffic, and many RARE members had to weigh the relative merits of keeping a network running against the desire to attract Commission money.

RIPE, of course, was the answer at first. RARE got in the act with the EBONE proposal, in which Kees took a leading role, along with groups like EUnet and NORDUnet. EBONE took the informal cooperation from 1991 and turned it into a more structured consortium for 1992.

Since the meeting I had attended in November at the Amsterdam Zoo, the EBONE proposal had taken shape. Two key groups, CERN and IBM, had decided not to sign the formal memorandum of understanding, but would still cooperate in the project.

Once the decisions to join (and how much to contribute) were made, the resources were turned over to an EBONE Action Team (EAT) to turn money, routers, people, and lines into a reasonable network.

A 512 kbps backbone had been decided upon between Stockholm, London, Amsterdam, Hamburg, and CERN. Three links to the U.S. were available, with two operating at 512 kbps and a T1 line from CERN (the EASInet IBM line). In some cases, money had been used to upgrade existing facilities. In others, existing resources were simply put into the EBONE pool.

The distinction was important because at the end of 1992, according to the Memorandum of Understanding that chartered EBONE, the group would disband and all resources would revert to the owners (except for the money of course, which is kind of hard to give back once you’ve spent it). No formal company existed and there was no dedicated EBONE manager.

At the end of 1992, the official plan was to have RARE start an Operational Unit. The Operational Unit would provide a home for things like EBONE and maybe even projects like IXI. The Operational Unit would be set up with capital from RARE members (and presumably some nice research grants from the Commission to keep it going). Officially, the Unit would operate as a separate organization, but with RARE members contributing the capital, RARE members would get the shares, and presumably, RARE would have some influence in how those shares got used.

Setting up the Operational Unit and operating it as a more formal replacement for EBONE could be looked at one of two ways. On the one hand, it could be the badly-needed professional operation that would run the long-heralded pan-European infrastructure. A few cynics looked at the Operational Unit in another way, seeing a power grab that could hurt existing operations such as EUnet.

The official raison d'être for the Operational Unit was contained in a report Kees handed me entitled “Final Report of the RARE Task Force on the Establishment of the Operational Unit for the Supply of Network and Information Services to the R&D Community” with a bright yellow cover and attractive GBC binding.

This report was a marvel of detail, containing a complete business plan for the Operational Unit with a cash flow analysis through 1996, job descriptions for OpUnit officers, and an annex of commonly asked questions and answers, my favorite being “Why are data networks important for Europe?”

This report explained how a professionally managed backbone was key for Europe. In fact, it was assumed that you needed to provide this capability through a single organization, sort of the way that the NSFNET had provided a single backbone service in the U.S.

The flaw in this analogy was that the NSFNET had not provided the single backbone service in the United States or for the Internet. The backbone was, as soon as multiple networks came into being, a combination of resources.

The NSFNET, for example, had to cooperate with the MILNET, the replacement for the ARPANET. NASA already had NSI, Energy had ESnet, and corporations had their own networks. A single backbone provider was certainly not the only way to structure European networking.

As I was looking through the business plan for the OpUnit, Kees handed me another document, a report from the European Engineering Planning Group, which in turn had been set up by the RARE Council of Administration.

The EEPG document had a grand plan for networking in Europe. Kees was a key member of the EEPG. The EEPG outlined a scheme in which research networking in Europe would need, to be a success, four bodies. An Operational Unit was one of them. A networking association (RARE), a policy body (people with money), and a “consultative networking forum” were the other three bodies.

Of course, there were other networking associations in Europe, but RARE had been doing a pretty good job of setting up agreements. RIPE had become a RARE activity, a status certainly placing it lower in the hierarchy than an association.

That left EARN. RARE had been engineering an acquisition and merger of EARN, proposing that the two groups form as a way of increasing their mutual power. Like a giant game of RISK, with Europe under control, RARE was taking an active role in the international scene, participating in places like the Coordinating Committee for Intercontinental Research Networking (CCIRN) and co-founding the Internet Society.

This wasn’t everything, either. I learned that RARE had helped form the European Workshop on Open Systems (EWOS) which would help set up European OSI profiles. RARE was also part of the European Telecommunications Standard Institute (ETSI), the official pan-European telecommunications standards body which was worrying about issues like ISDN or 2 Mbps X.25 networks and had joined the European Council of Telecommunications User Associations (ECTUA).

I left Kees' office with my head swimming. He had let me know that he had to leave by 4:06. I had thought the departure time was a bit odd, but it occurred to me that Kees was in the shopping center of which the train station was a part and that he had probably timed the walk down to the trains. Just in case he felt like getting a jump on things, I left at 4:03.

Walking back to town, to keep warm, I went through all the new acronyms I had learned, hoping that repeating them would generate enough hot air to counteract the chill. Needing a cup of coffee and a blank piece of paper to sort everything out, I ducked in out of the cold into a little corner coffee shop with windows all fogged up a definitely counterculture clientele.

Rubbing my hands to warm up, I stepped up to the counter to order.

“Do you have a menu?” I asked the attendant who ambled over from a table where he was playing chess.

He pointed down on the glass. I quickly realized that I was at the wrong counter. This menu specialized in regional specialties. The left hand side featured different kinds of hashish, ranging from your basic Lebanese to your more exotic Nepalese. The 10 and 25 guilder columns contained how many grams (or portions of grams) you got for your money. The right hand side of the menu was devoted to fine marijuanas, featuring orange bud, sinsemilla, and the like.

Not needing to confuse myself even further, I went down the steep spiral stairs to the coffee bar in the basement. There, amidst the kids rolling joints, the din of pinball, and the very loud music, I read through the stack of documents I had collected that day.