Summary

  • A mobile packet-radio node sent traffic through gateways and three different packet networks toward hosts in the United States and Europe.
  • The demonstration showed that a common internetwork protocol could work across the tested route; it did not prove universal deployment, routine reliability or a finished TCP/IP Internet.

The van was the endpoint, not the accomplishment

The image is easy to remember: a research van moving along California roads, its antenna linking it to a radio network. But a moving vehicle alone could not answer the question that mattered. By 1976, researchers had already bridged the Packet Radio Network and the ARPANET. The next test had to show whether a third, differently engineered network could join the path without being rebuilt as part of the first.

Vint Cerf later described the 1977 demonstration as a way to show ARPA director George Heilmeier that multiple networks could appear to operate as one uniform internet. In his oral account, a vehicle on the Bay Shore Freeway reached packet-radio repeaters around the Bay Area, crossed a gateway into the ARPANET, traversed the packet-satellite network to University College London, used a dedicated satellite link associated with Norway, returned to the ARPANET and reached a host at the University of Southern California. The path was an engineered chain, not a magic, borderless cloud.

The three networks were distinct operating environments: mobile packet radio, the wired ARPANET and SATNET. Gateways sat at the joins. BBN supplied gateway software; SRI’s people and equipment put the mobile radio segment on the path; university and research teams supplied hosts and network operations. Cerf recalled that the networks were not to be changed simply to make them look alike. The common protocol had to carry communication across their differences.

What one completed path establishes

The demonstration’s strongest result was concrete: data traversed a planned route through three packet networks and reached a remote host. That was more than a drawing or a protocol paper. It was evidence that separately operated network technologies could participate in one end-to-end communication when hosts and gateways implemented compatible rules.

It was less than a census of readiness. The public accounts do not provide a broad matrix of repeated tests, traffic loads, packet loss, recovery time or independent operators. A carefully prepared route with installed gateways cannot show that arbitrary networks would interoperate, that every packet arrived under every condition, or that a service was ready for ordinary users. Nor should the 1977 protocol be casually relabelled as the complete TCP/IP architecture later deployed at scale.

The date also deserves care. RFC 2235 lists a first demonstration of ARPANET, Packet Radio Net and SATNET in July 1977. The Computer History Museum’s anniversary account identifies the major transmission as 22 November. These may refer to different stages, but the cited records do not settle that point. The sound claim is that the three-network demonstration took place in 1977; the day is not secure across the sources.

A milestone measured at the boundary

Read through a later design lens, the experiment illustrates two ideas in Heng Lu’s Notes 64 and 65: keep the shared specification to what interoperability requires, and judge a protocol by running implementations. That is an interpretation applied decades later, not the participants’ documented terminology. The historical evidence is simpler: the gateways and hosts worked together on this route while the underlying networks remained different.

That distinction keeps the milestone useful. The test did not make every network part of one centrally operated system. It showed that a common communication rule could cross boundaries whose local technologies and operators remained separate. Its achievement was not universality. It was a path that worked—and a practical reason to try the next one.

Sources