Summary

  • ARPA Order 471, dated 5 April 1963, funded two distinct three-year research plans: UCLA's computer-network and time-sharing work and Berkeley's remote stations and programs for a computer network.
  • The plans treated interconnection as a set of open research problems. The order is evidence of an early funded effort, not proof that a working network resulted or that Order 471 became ARPANET.

Analysis

The familiar origin story often moves quickly from J. C. R. Licklider's “Intergalactic Computer Network” to ARPANET. An April 1963 funding order sits in the space between those two kinds of evidence. It records what the Advanced Research Projects Agency (ARPA) was prepared to pay for at that moment: research on computer networks, time-sharing and remote stations under its Command and Control Research program. It does not record an Internet already waiting to be built.

The order was dated 5 April 1963. Its first page requests two information-processing contracts, each for three years. One assigned computer-network and time-sharing research to the University of California, Los Angeles. The other assigned remote stations and programs for computer networks to the University of California, Berkeley. The attached Program Plan 93 gives UCLA an initial three-year funding figure of $675,000. Program Plan 95 gives Berkeley a three-year level of effort of $797,000.

Those figures belong to the proposed contracts in the surviving order; they are not a complete accounting of every later expenditure associated with the wider effort.

The plans were grounded in particular machines and institutions. UCLA's Western Data Processing Center already served UCLA and more than thirty participating institutions, with teleprocessing links to six. ARPA proposed to connect that environment to the AN/FSQ-32 at the System Development Corporation in Santa Monica. Berkeley's plan called for a remote station and a data link to the same Q-32, while augmenting its IBM 7090 and adding a smaller satellite computer. The proposal also left a patch-panel provision for a possible future link to Stanford.

This was a research arrangement whose sites, machines and contracts were named; it was not yet a general-purpose network service.

The technical list is more revealing than the later nickname. Plan 93 names dynamic memory allocation, language and format compatibility, common files that different operators might change, limited-capacity communications channels, priority assignment, record keeping and incremental compilation. Plan 95 adds remote control, display stations and the preparation of large programs while the computers were occupied with other work. The problem was not simply how to send bits between two buildings. It was how to let separately administered machines, software, users and work schedules cooperate without pretending that they were one computer.

The institutional case was equally explicit. ARPA linked the work to military command-and-control systems expected to involve computer-to-computer communication. Yet the order selected university environments because their existing facilities and research groups could supply capabilities relevant to that mission. UCLA's operating service network helped justify its role; Berkeley's computer center and researchers in programming languages and heuristic programming justified another. ARPA also expected coordination among contractors, visits and an informal technical group.

Shared conventions were to emerge where they helped; the plans did not prescribe one complete, centrally imposed system.

Licklider's memorandum of 23 April 1963 to “Members and Affiliates of the Intergalactic Computer Network” helps explain the coordination problem, but it is a different document from Order 471. The memo calls the enterprise difficult to name and its purpose not yet clearly structured. It asks participants to compare projects, dependencies, costs and opportunities for mutual benefit before plans harden. It raises the possibility of shared conventions while warning against rigid rules designed merely to maximize interchangeability.

It also asks who would control a network language, how one center would interpret another's request, and how shared files would be protected while they changed. The memo is evidence of Licklider's questions to a research community; it is not the specification of the two contracts.

The record also places a boundary around what can be claimed. The National Archives' retrospective account of the ARPA Order files says Order 471 ended in 1970 and that the documentation available to its reviewers showed no results, after more than $5 million in expenditure. That is an archival statement about the reviewed project folder and the limits of its available documentation. It does not prove that no useful work occurred, that no report or technique influenced anyone, or that the money was spent under the two April 1963 plans alone. The attached plans themselves describe intended work, not completed results.

For that reason Order 471 is best read as a funded attempt to make distributed computing administratively and technically workable, not as a miniature ARPANET. Its evidence shows research contracts, computer centers, proposed links, work packages and a long list of problems the participants still had to solve. Later ARPANET history belongs to a separate chain of decisions and implementation records. Keeping the two apart makes the early story less tidy but more informative: networks required budgets and institutions as well as ideas, and a plan's existence cannot substitute for a receipt showing what ran.

Sources