Packet switching

In August 1964, Paul Baran at the RAND Corporation published an eleven-volume analysis titled 'On Distributed Communications' for the U.S. Air Force. The military wanted a communications network that would survive nuclear attack—one where the destruction of any node would not bring down the system. Baran's solution was radical: break messages into small 'message blocks,' send each block independently through a mesh network using 'hot-potato routing' where nodes pass packets toward their destination like players passing a hot potato, and reassemble the blocks at the receiving end. No central control. No single point of failure. Information would find its way like water flowing downhill, routing around damage.

Unknown to Baran, British computer scientist Donald Davies at the National Physical Laboratory independently developed the same concept in 1965. Davies called his units 'packets'—a term that stuck because it translated well into other languages. Davies designed a complete store-and-forward system optimized for interactive computing, envisioning a network where terminals could share remote computing resources. When Davies' colleague Roger Scantlebury presented this work at a 1967 conference, the ARPA team building a computer network took notice.

The adjacent possible for packet switching required recognizing that digital communication could work fundamentally differently from telephone circuits. Telephone networks established dedicated circuits between callers—if the circuit broke, the call failed. Baran and Davies independently realized that digital messages could be broken apart, routed through multiple paths, and reassembled—requiring no dedicated circuit and surviving partial network failure. This insight drew on queuing theory, computer time-sharing concepts, and the military's experience with survivable command structures.

Why did two researchers on opposite sides of the Atlantic arrive at the same solution simultaneously? The Cold War created identical pressures. Both the United States and Britain needed communications that could survive attack. Both had access to digital computers capable of the store-and-forward processing that packet switching required. Both were thinking about computer resource sharing. The adjacent possible had opened; Baran and Davies walked through independently.

Baran remained gracious about the parallel invention. Decades later, he wrote to Davies: 'You and I share a common view of what packet switching is all about, since you and I independently came up with the same ingredients.' Larry Roberts, who built ARPANET, acknowledged that the networks of the 1970s were similar 'in nearly all respects' to Davies' original 1965 design.

ARPANET deployed in 1969 with its first node at UCLA. The network that would become the Internet used packet switching as its foundation. Every email, every web page, every video stream today is broken into packets, routed independently through the network, and reassembled at the destination—exactly as Baran and Davies envisioned sixty years ago. The architecture that was designed to survive nuclear war proved equally suited to surviving the chaos of a global network of networks, where paths fail and recover constantly, and information finds its way regardless.