Summary

  • NSPIXP-1 was hosted at WIDE’s Tokyo NOC, but its exchange segment was configured outside the WIDE Internet AUP.
  • A 1995 WIDE report records an October 1994 move from Layer 3 interconnection toward Layer 2, shifting routing-policy control toward participating providers.

A boundary inside the building

The telling detail about NSPIXP-1 is not simply where it was installed. The first Japanese network service provider Internet exchange occupied space at WIDE Project’s Tokyo network operations centre, WNOC-TYO, while its exchange segment was configured as independent of the WIDE Internet and outside the scope of WIDE’s acceptable-use policy. Physical hosting and policy jurisdiction did not have to coincide.

That separation mattered because an exchange joins networks that have their own operators, customers and traffic decisions. A research network could provide a practical site without making every connected provider part of its own network or subject to its own use rules. WIDE’s 1995 annual report describes the separation as part of the experiment’s design; it does not establish that every participant saw the arrangement as neutral, or that the same boundary solved every interconnection problem. WIDE Project, 1995 report

The report places the work in two phases: Phase I, April 1994 to March 1995, and Phase II, April 1995 to March 1997. Those are the report’s project periods, not a single launch date. JPNIC’s chronology distinguishes installation in March 1994 from operations beginning in December, a useful reminder that equipment, experiment and service do not share one timestamp. JPNIC timeline

The routing architecture changed during Phase I. WIDE says the initial design used Layer 3 interconnection because providers had unequal technical capabilities. In October 1994 the project moved toward Layer 2. In the report’s account, the Layer 3 arrangement left routing policy at the exchange and made providers dependent on that policy; the Layer 2 design let the participating providers operate their own routers and make routing choices. A later WIDE retrospective explains the same distinction as a shift in who chose policy: the exchange operator in Layer 3, each ISP in Layer 2. WIDE Project, 2005 retrospective

The change did not remove coordination work. As the number of providers and BGP configurations grew, Phase II examined route registries and route servers. Those tools addressed a practical question: how could separate networks exchange enough routing information to connect without turning the exchange operator into the permanent author of each provider’s routing policy?

The commercial pressure came from the networks themselves. In a 2011 JPNIC history, WIDE participant Osamu Nakamura recalls that providers wanted domestic traffic to avoid costly passage through the United States and worried that inbound traffic would consume their own resources. His account describes an early “fuse” model in which each provider chose the capacity of its own exchange line; it gives 192 Kbps against a 1.5 Mbps T1 backbone as an example. Nakamura says four providers joined at first, the count rose above twenty by the end of 1995, and later fuse saturation led to packet loss. These are retrospective recollections, not an independently measured traffic series. Nakamura’s JPNIC history

NSPIXP-1 matters because its record makes two often blurred questions visible: who hosts an exchange, and who controls the network policy applied to it? In this case the host and the policy boundary were deliberately separated, then the routing model moved toward decisions made by the networks operating at the edge. The linked WIDE Project directory entry supplies the institutional context; this article concerns a particular experiment, not a new profile of WIDE.

The evidence supports a bounded conclusion. It shows an intended AUP separation, a documented change in interconnection model, and later participant recollections about capacity pressure. It does not show a complete route-by-route history, prove that all domestic Japanese traffic stopped crossing the United States, or establish that every provider adopted identical practices.