Summary

  • Amsterdam’s exchange operated as shared infrastructure before its participants founded a separate association in 1997.
  • The later member vote changed who governed the service; it did not make every network a peer or erase the commercial terms required to connect.

The February 1996 announcement read like an equipment plan. AMS-IX was introducing switched Ethernet and FDDI, while placing its Layer 2 equipment at SARA. The message also described the access rule: an Internet service provider could connect if it already had an arrangement to exchange traffic with at least one network on the exchange. That combination matters. A shared switch could lower the practical barrier to interconnection, but it could not make two networks agree to exchange routes or traffic. The announcement is preserved in the RIPE mailing-list archive.

The exchange itself was older than the 1996 upgrade. The same notice says its predecessor, the International Backbone Router LAN, had been in use since the end of 1993 and ran through the computer rooms of SARA, CWI and NIKHEF. It calls February 1994 the date AMS-IX was formally established. These dates describe stages: a local interconnection fabric first, a named exchange next, then a distinct legal and governance structure. AMS-IX’s later 2019 annual report likewise traces the system to Amsterdam’s research centres and the commercial networks that arrived near them after the Dutch telecommunications market opened in 1994.

That geography joined different jobs. SARA and NIKHEF hosted the exchange infrastructure; CWI had earlier served as an important European gateway to the US network. CWI’s 2003 retrospective says it handed that gateway role off in 1996 and describes AMS-IX as located with its Science Park neighbours. The distinction is easy to blur: an international gateway carries traffic between networks and external paths, while an exchange lets connected networks interconnect locally. The SURF history records AMS-IX as a pilot project in 1994, inside a wider research-network environment. The available chronology does not support treating the exchange as a single replacement for every Dutch or European backbone function.

By February 1996, the participants named in the RIPE announcement included Ebone, EuropaNET, EUnet, NLnet and SURFnet. Adding switched Ethernet and FDDI expanded the physical ways to attach networks. It did not abolish bilateral choices: the connecting ISP still needed an exchange arrangement with an existing participant. Nor does a common Layer 2 fabric prove that every connected network exchanged traffic with every other one. The switch supplied a common meeting surface; each participant’s policy and commercial agreement still determined which relationships existed.

The governance change followed the operational one. AMS-IX’s 2019 report says participants held regular technical meetings from 1994 to 1997. In 1996, they decided to establish a structure independent of SURFnet. On 1 January 1997, they founded a neutral, non-profit association. Connected parties became members and received votes in general assemblies. The association did not invent the exchange; it gave the networks already using it a formal role in governing its shared service.

A later snapshot shows the difference between a local pilot and a mature exchange. At RIPE 31 in September 1998, participants reported two host organisations, SARA and NIKHEF; switches linked over private fibre; and 40 connected networks, 24 international. The RIPE minutes also recorded substantial joining and annual fees. Those figures are from 1998, not the founding year. They show both growth and a limit to the word “open”: a member vote can broaden accountability while price and eligibility still shape who gets a port.

The founding perspective is visible in Henk Steenman’s AMS-IX interview: he took part in establishing the association and served on its board from 1997 to 2001. AMS-IX’s 30-year account names NIKHEF, SARA, CWI, SURFnet and NLnet among the parties involved. Neither retrospect turns the exchange into a one-person invention. Its sequence is more useful: a running facility acquired a broader technical base, then its connected participants assumed formal governance.

AMS-IX’s early history is therefore not a choice between a research network and a commercial one. The commercial providers joined a facility whose locations, equipment and operating expertise had grown from research institutions. The 1997 association changed the control surface while preserving a service that already connected networks. That handoff mattered because the Internet’s local interconnection layer needed a durable operator, while the operator did not have to remain the only institution with a voice.