Summary

  • TAT-8’s planned branching unit made the choice of European landings a physical and commercial decision: the design connected New Jersey to Britain and France, while a proposed Spanish branch required agreement from every co-owner.
  • Its optical capacity was not an open Internet commons. Owners reserved capacity against forecasts, and non-owners depended on excess capacity or a consented lease.

A fork in the seabed fixed the map

A cable route can look inevitable once it is drawn on a map. TAT-8’s 1984 plan shows otherwise. A Canadian government study described a transatlantic trunk from Tuckerton, New Jersey, to an underwater branching unit near Europe, then separate legs to Widemouth in England and Penmarch in France. The consortium had declined proposed spurs to Canada, Spain and Portugal. The plan left room to add Spain later, but only if every co-owner agreed.

That branching unit was not a packet router. It divided a physical transmission system and the capacity attached to each path. The distinction matters: Internet protocols can select among routes that operators have made available, but they cannot create a landing point or a fiber pair that the cable’s owners did not build and allocate. In TAT-8, the geography came first.

One Atlantic system, three suppliers

The same 1984 study assigned the long U.S. trunk to AT&T, the British leg to STC and the French leg to Submarcom, with AT&T responsible for system-wide compatibility. Robert Paski’s 1989 engineering paper describes the owners’ decision to split TAT-8 into three separately supplied subsegments after three companies bid for the system. The cable was therefore both one route and a negotiated boundary between suppliers.

That arrangement distributed construction work across firms with national and technical stakes, while creating an integration job at the joins. A connected optical path had to behave as one communications system even though no single supplier made every segment. This is a different kind of coordination from an Internet standard: matching physical equipment and service contracts must precede any higher-layer choice of protocol.

Digital did not mean “the Internet”

TAT-8 entered service in December 1988 as a general digital communications cable. Contemporary reporting described voice, computer and video traffic; the 1989 engineering paper calls it the first digital fiber-optic transoceanic cable system. Neither fact makes the cable an Internet protocol or proves which later IP circuits used it. “Digital” describes the transmission system, not the application or the network-layer agreement riding over a leased circuit.

CERN’s 1989 annual report offers a careful point of comparison. It records a 1.5 Mbit/s connection from NSFnet via Cornell, donated by IBM, as on order. A later CERN history places that link in the growth of EASInet and the European Internet. The annual report does not name the submarine cable carrying the circuit, so this article does not label it a TAT-8 link. The evidence shows that high-speed academic interconnection was being arranged in the same period; it does not identify the precise seabed route.

Capacity had owners before it had users

The 1984 FCC order treated capacity as an asset to be assigned among co-owners. Their shares reflected forecast requirements for 1988–95. The application did not propose a common pool of U.S. capacity for later allocation: non-owners could seek capacity beyond the owners’ aggregate need, while transferring or leasing jointly held capacity required the relevant joint holder’s consent. A cable could be shared without being open by default.

That distinction is a useful way to read TAT-8 in Internet history. Protocols may allow independent systems to interoperate without one central traffic controller. The path beneath them still depends on capital, landing permissions, supplier interfaces and capacity rights. TAT-8 did not invent the Internet’s decentralized architecture; it makes visible some of the concentrated infrastructure choices that architecture could not erase.

Primary sources