Summary

  • Kees Neggers helped initiate Ebone after a 1991 SURFnet multiprotocol pilot showed, in that configuration, that a nominal 2 Mbps X.25 service could not carry more than a reported 360 Kbps.
  • Ebone was not only a set of circuits. Contributing organisations, an action team, a management committee, a Stockholm operations centre and RARE’s clearing-house role made an interim network governable while European institutions formally favoured OSI.
  • Its completion in 390 days supplied practical evidence for a multiprotocol policy, but the record does not support a lone-founder story or the claim that one backbone by itself ended Europe’s protocol transition.

The last link

The message to the European IP community was sent late on a Saturday. The London–Montpellier link was in place; the initially defined Ebone was complete. Its author called the result a resilient pan-European IP backbone and looked back to 1 September 1991, when a small group had gathered in an office building outside Amsterdam. The elapsed time was 390 days.

That contemporary note is useful because it gives the achievement a physical edge and a social perimeter. There was a final circuit, but there had also been a table of people, voluntary labour and organisations willing to contribute. The note thanked a group with different kinds of expertise. It did not present a solitary inventor throwing a switch.

Kees Neggers belonged at that coordination table. Internet Society’s institutional biography identifies him as the initiator of the 1991 Ebone effort and chair of the Ebone Consortium from 1991 to 1994. By then he had worked in international research networking since 1984, served in the leadership of RARE and managed SURFnet from its creation in 1988. Those posts gave him no magical power over European networks. They put him at the interfaces where national networks, standards bodies, public carriers and research users had to agree enough to operate something together.

The pilot that made a limitation visible

The immediate problem was not that Europe had no networking. It had too many systems that could not yet be treated as one service. Research communities had built discipline-specific networks. National organisations were developing their own infrastructures. EARN connected compatible systems; European policy invested heavily in OSI; X.25 supplied a familiar carrier service; and IP was already carrying work that users did not want to lose.

SURF’s own history captures the layered transition. SURFnet bv was established on 1 January 1988 by SURF and the Dutch PTT. Its aim was not to add one more island beside the local networks of universities and laboratories. It wanted a transparent national network that could incorporate them. The first production version, SURFnet2, went live in 1990 with IP transport alongside X.25 and a stated maximum capacity of 64 kbit/s.

In Neggers’s later account, SURFnet and the PTT ran a 2 Mbps multiprotocol pilot in 1991. The X.25 implementation in that test could not support more than 360 Kbps, he said. That figure needs a narrow label. It is not a universal law about X.25, nor a controlled benchmark preserved here with packet sizes, equipment and test logs. It is the reported observation that turned an architectural concern into an operating constraint for the people deciding what to build next.

The test changed the quality of the disagreement. “Which protocol ought Europe to endorse?” could remain a contest of specifications and institutional commitments. “Which configuration carries the work at the required rate?” could be examined by operators and users. The pilot did not abolish politics. It gave politics a result it had to explain.

A backbone with organisational interfaces

The next step was a task force for a European multiprotocol backbone. RIPE’s record of its twelfth meeting shows how much institutional machinery sat behind the word “backbone”. Interested organisations formed the European Consortium of Contributing Organizations, or ECCO. That group designated an Ebone Management Committee and an Ebone Action Team. Implementation began in the first part of 1992. The Royal Institute of Technology in Stockholm was accepted as the network operations centre.

Each role answered a different failure question. Contributing organisations had to make resources available. A management body had to settle scope and continuity. An action team had to turn plans into configurations and installed links. A network operations centre had to see incidents that crossed national boundaries. In Neggers’s retrospective, RARE supplied administrative support and became the clearing house. Accounting, contribution and responsibility were therefore not paperwork attached to the network after it worked; they were part of the mechanism that allowed it to keep working.

The topology also embodied an argument. The completion message names London and Montpellier as the last link of an initially defined system. RIPE minutes describe Stockholm, Amsterdam and CERN among the kernel sites and treat Ebone as complementary to other European backbone initiatives. It was not necessary for one programme to erase every other one before packets could move. The interim network could interconnect contributions and coexist with the services whose limitations it was exposing.

This is the practical meaning of an open exchange in the Ebone story. Openness was not the absence of rules. It was a rule about the boundary: different networks could meet through an interface without surrendering their entire internal design to a single owner. The Internet Hall of Fame credits Neggers with advancing that model. The stronger historical claim, however, belongs to the operating record itself: a multi-organisation system acquired a topology, a NOC, working groups and a clearing function quickly enough to become evidence.

Running beside the official track

It is tempting to tell the period as insurgent IP defeating a forbidden official network. The sources do not support that drama. Neggers recalled that governments and public telecommunications operators formally backed OSI initiatives, but also understood the need for Ebone and did not forbid it. Ebone was presented as an interim multiprotocol activity and even included a pilot connectionless OSI service.

The distinction matters. A tolerated parallel service can learn without forcing every sponsor to renounce its existing policy on day one. It lowers the institutional cost of discovering that the official plan needs revision. At the same time, calling an activity “interim” can conceal power: the service that attracts users, builds operational competence and accumulates working routes may become the default before a formal vote catches up.

According to the GÉANT retrospective, Ebone’s success was followed within a year by COSINE and European Commission approval for the next generation of European backbones to be multiprotocol. That sequence supports a modest conclusion. Operating evidence entered the formal decision. It does not isolate Ebone as the sole cause. NORDUnet’s choices, user demand, equipment performance, carrier economics, other pilots and the growing global use of IP all belonged to the transition.

Even the phrase “protocol wars” can flatten what operators faced. They were not choosing coloured flags. They had installed systems, contracts, skills, addressing practices and users whose work crossed several architectures. A useful transition had to preserve service while changing the basis of service. Ebone’s contribution was to make one route through that transition observable.

Completion was a checkpoint

The 390th day did not settle every question. “Complete” referred to the initially defined backbone, not to the end of Ebone’s development, Europe’s network build-out or the commercial history that followed. The surviving material in this package does not provide an audited series for availability, delay, packet loss, security incidents or user outcomes. Resilience in the completion announcement describes the intended topology; it is not a substitute for measurements that are not present.

Nor should later institutions be poured backwards into Ebone. RIPE NCC developed its own legal and operational identity. AMS-IX began as a separate exchange at Amsterdam’s Science Park. NetherLight and GLIF addressed an era in which optical capacity and data-intensive science created different coordination problems. Eduroam arose from another proposal and another team. Neggers participated in or encouraged each of these environments, but institutional continuity is not organisational identity.

There is still a defensible pattern. The Internet Hall of Fame describes Neggers as an advocate of open, inclusive and bottom-up self-regulation. In 2001, SURFnet put 10 Gbit/s IP directly over optical waves and began the NetherLight Open Lightpath Exchange. A 2004 GLIF release describes international network managers and scientists treating wavelengths—and therefore bandwidth itself—as schedulable resources.

The scarce object changed from a shared backbone route to an optical path, yet the operating question remained: who can contribute capacity, who can request it, and which interface lets independent parties coordinate without becoming one organisation?

That pattern is more informative than a list of boards. It treats institution-building as technical work with failure modes. A clearing house can fail to reconcile contributions. A NOC can see links but miss a user community. An open exchange can become dependent on an opaque operator. A provisional architecture can become permanent without an explicit review. Conversely, an institution can protect experimentation by giving a temporary service named responsibilities, observable performance and a path into formal decisions.

Kees Neggers’s Ebone record is therefore not simply a story about choosing IP. It is a story about making a choice testable. Europe did not have to wait for every standards argument to finish before learning from a network. It built one in parallel, gave the experiment operational custodians, and let the result confront the plan.

Sources