ARPANET

J.C.R. Licklider (1915–1990), psychologist and computing visionary at ARPA, conceived the "Galactic Network" concept in 1962—a globally interconnected set of computers enabling human-machine symbiosis. Though he did not build ARPANET itself, his intellectual framework and leadership of ARPA's Information Processing Techniques Office (1962–1968) made the network's existence possible. Bob Taylor, who succeeded Licklider, secured funding and championed the project; Larry Roberts directed its technical design; and Vint Cerf and Bob Kahn later architected TCP/IP, the protocols that transformed ARPANET into the modern Internet.

ARPANET's architecture rejected the circuit-switched model of the telephone system in favor of packet switching—a method developed independently by Paul Baran (RAND Corporation, 1960s) and Donald Davies (National Physical Laboratory, UK, 1965). Data was fragmented into small packets, each routed independently across the network via the most available path, then reassembled at destination. This design conferred resilience: the loss of any single node or link did not cripple the network. Interface Message Processors (IMPs)—ruggedized Honeywell minicomputers running specialized software—served as packet routers at each node. Host computers connected to IMPs via a standard interface, allowing heterogeneous machines (IBM mainframes, DEC PDP-10s, etc.) to interoperate. The Network Control Protocol (NCP), deployed in 1970, governed host-to-host communication; it was later superseded by the more flexible and scalable Transmission Control Protocol/Internet Protocol (TCP/IP) suite, adopted network-wide in 1983.

In 1962, J.C.R. Licklider, newly appointed head of ARPA's Information Processing Techniques Office, articulated a vision of networked computing that would enable interactive, real-time collaboration among researchers across the nation. The Cold War context was crucial: ARPA (renamed DARPA in 1972) sought to build a communication system robust enough to survive nuclear attack—one without a central hub that could be destroyed. By 1966, Bob Taylor had secured $1 million in funding; Larry Roberts was hired to design the network. Roberts adopted packet switching, a radical departure from AT&T's circuit-switched telephone infrastructure. On September 2, 1969, the first IMP was installed at UCLA; by December 1969, four nodes were connected. ARPANET grew steadily: 10 nodes by 1970, 23 by 1971, 57 by 1974. The network remained largely confined to academic and military research communities until the 1980s. In 1983, TCP/IP became the mandatory protocol suite, unifying ARPANET with other emerging networks (CSNET, NSFNET). The transition from ARPANET to the Internet was gradual; ARPANET was formally decommissioned in 1990, but its architecture, ethos, and protocols live on in the modern Internet.

ARPANET was born from three converging imperatives: (1) Cold War strategy—the U.S. Department of Defense needed a communication network that could survive nuclear war by having no single point of failure; (2) Scientific efficiency—researchers at universities and labs across the country needed to share expensive computing resources and collaborate in real time; (3) Technological vision—Licklider and his peers believed that human-computer interaction, augmented by networked communication, would amplify human intellectual capability. The project embodied the belief that information technology could be a force multiplier for American science and defense. Funding flowed because ARPA had both the mandate and the budget to pursue high-risk, high-reward research.

ARPANET users—primarily computer scientists, engineers, and military researchers—employed the network for file transfer (via FTP, deployed 1971), remote login (via Telnet, 1969), and electronic mail (Ray Tomlinson's email system, 1971). A researcher at Stanford Research Institute could log into a computer at UCLA, transfer data, and collaborate with colleagues in real time—a revolutionary capability. By the late 1970s, ARPANET traffic included not only data and files but also early mailing lists and discussion forums. The network was not user-friendly by modern standards; interaction required command-line interfaces and technical knowledge. Nevertheless, ARPANET users discovered the social and intellectual power of networked communication, laying the groundwork for the Internet's later explosion.

ARPANET was not a physical object but a distributed system assembled over time. The first phase (1968–1969) involved the design of the IMP and the NCP protocol. BBN Technologies, a Cambridge, Massachusetts research firm, won the contract to build the IMPs. The Honeywell H316 minicomputer was selected as the IMP platform; BBN engineers wrote the packet-switching software in assembly language. Leased telephone lines (50 kbps) were rented from AT&T to connect nodes. The first IMP arrived at UCLA on August 30, 1969; installation and testing took two weeks. On September 2, 1969, the first packet was transmitted from UCLA to Stanford Research Institute. Subsequent nodes were added throughout 1970 and 1971. The network grew organically, with each new node requiring a new IMP, new leased lines, and integration with existing infrastructure. In the early 1980s, the transition to TCP/IP required software updates across all nodes but no major hardware replacement. The network's decentralized, modular design allowed incremental growth without redesign.

ARPANET itself was a singular network, but it spawned several related systems: CSNET (Computer Science Network, 1981–1991), funded by the National Science Foundation, extended networking to universities without ARPANET access. NSFNET (1985–1995), also NSF-funded, provided backbone connectivity for academic institutions and later interconnected with ARPANET. MILNET (Military Network, 1983), a spinoff of ARPANET, served military and defense research communities. International variants emerged: the British JANET (Joint Academic Network, 1984) and the European EARN (European Academic and Research Network, 1982) adopted similar packet-switched architectures. These networks eventually interconnected via TCP/IP gateways, forming the Internet. The key variation was not in ARPANET's core design but in the protocols layered atop it: NCP (1970–1983) gave way to TCP/IP, which proved far more scalable and flexible.

ARPANET itself was the famous example—the first operational packet-switched network. Key nodes included UCLA (the first), Stanford Research Institute (the second), UC Santa Barbara, and the University of Utah. The network's most celebrated moment came on September 2, 1969, with the first packet transmission. By the mid-1970s, ARPANET had become the de facto standard for academic and military research networking in the United States. The network's most famous application was email, which emerged organically from the need to send messages across the network and became the killer app that drove adoption. The ARPANET mailing list (created in 1979) was an early form of online community, allowing researchers to discuss technical topics asynchronously. The transition to TCP/IP in 1983 was a watershed moment, as it unified ARPANET with other emerging networks and set the stage for the modern Internet.

ARPANET was not a physical artifact in the traditional sense, but its material legacy survives in museums and archives. The Smithsonian Institution holds artifacts related to early computing and networking, including documentation of ARPANET's development. The Computer History Museum in Mountain View, California, preserves hardware and software from ARPANET, including IMP documentation and early network protocols. BBN Technologies' archives contain engineering records, software listings, and correspondence related to IMP development. The Internet Society maintains historical records of ARPANET's evolution, including RFC (Request for Comments) documents that chronicle the network's technical development. Individual universities that hosted early ARPANET nodes (UCLA, Stanford, UC Santa Barbara, University of Utah) have archived their own histories and technical documentation. No physical IMP survives in operational condition, but photographs, schematics, and software source code document the machines' design and function.