Zum Inhalt springen

The Myth: The Internet Was Designed to Survive Nuclear War

Zusammenfassung

The most persistent myth about the internet’s origins is that ARPANET was designed to survive a nuclear attack. It was not. ARPANET was designed for resource sharing between university researchers — allowing a computer at UCLA to use computational resources at MIT. The nuclear-survivability design that inspired the myth came from a separate RAND Corporation study by Paul Baran, published in 1964, which described a distributed communication network that could survive partial destruction. ARPANET’s engineers knew of Baran’s work but did not build ARPANET to his specifications. The myth persists because it makes the internet sound more dramatic than “let’s share computers between universities.”

The Actual Origin

J.C.R. Licklider at ARPA articulated the resource-sharing vision in his 1963 “Intergalactic Computer Network” memo. The problem: the computing facilities at different universities (MIT, CMU, Stanford) were inaccessible from remote locations. Researchers traveling between institutions couldn’t use their home institution’s software and data. Licklider’s memo proposed connecting these machines so that a researcher could access computing resources from anywhere.

Larry Roberts implemented this vision starting in 1966. The network that BBN built — packet-switched, with Interface Message Processors at each node — was designed to be reliable, interoperable, and expandable. Packet switching (routing individual packets independently, allowing them to take different paths to the same destination) made the network robust to individual link failures. But robust against link failures is not the same as designed for nuclear survivability.

The Baran Study

In 1964, Paul Baran at RAND Corporation published an 11-volume study titled “On Distributed Communications.” Baran’s study was genuinely about surviving a nuclear attack: he analyzed what kind of communication network would continue to function if a significant fraction of its nodes and links were destroyed. His answer was a highly redundant, distributed mesh network where every node could route packets around failed sections.

ARPA was aware of Baran’s work. Roberts cited it in early ARPANET papers. Some of the packet-switching concepts in ARPANET drew on Baran’s analysis. But ARPANET was not built to Baran’s redundancy specifications, was not funded as a military communication backup, and was not designed with nuclear scenarios as the primary use case.

The ARPANET security classification was minimal: it was a research network accessible to university computers with relatively loose access controls. A network designed to survive nuclear war would not have had those characteristics.

Why the Myth Persists

The nuclear-survivability story is more compelling than the resource-sharing story. “Designed to survive the bomb” implies strategic importance, dramatic engineering, and Cold War relevance. “Designed to let UCLA borrow Stanford’s computer” does not have the same narrative force.

The myth also has a kernel of technical truth: distributed packet switching is more robust than centralized circuit switching, and this robustness was a desirable property for military communication networks. ARPANET’s architecture, if not its specifications, would have been a reasonable starting point for a survivability-focused network.

The full origin story is in ARPANET: Building the Network and The Connected World.

📚 Sources