A Non-Hierarchical Network
On New Years Eve, 1960, Paul Baran published his historic memo titled On a Distributed Command and Control System Configuration. The document sketched out the strategic implications of a set of network topologies, and also proposed the basic technical concept of packet-switching which now underpins how we use network technologies. Baran's memo predates ARPA's work on networking by about five or six years, and it was approximately contemporaneous to J.C.R. Licklider's paper Man-Computer Symbiosis (1960).
Although ARPA's origin and much of its research was firmly in the military domain and primarily in advanced weapons and defence systems, the ARPANET was first and foremost designed as a practical tool for coordinating and sharing research amongst scientists and engineers working across the United States. Revising the popular assumption, many scholars are keen to point out that the ARPANET was not, at least in its initial design and development, a military system. Paul Baran's design however was, although perhaps not in the way we might expect. To explain the strategic value of his proposal, some technical explanation is necessary.
This image, featuring two figures of network topologies in the On a Distributed Command and Control System memo, offers a useful point of comparison to explain precisely what Baran was aiming at. The top diagram is what he refers to elsewhere as a "pure hierarchical network", whereby information from a set of nodes (A, B, C etc) have to pass through other bottleneck nodes (A*, B*, C* etc) and centralised points (A**, B**, C**) in order to travel across the wider network. For an attacked wishing to bomb and neutralise the infrastructure, this poses a logical optimal strategy. He writes: "This of courses places a high premium upon the destruction of stations A*, A**, B*, etc. We are interested in other network configurations that minimise this attacker's bonus" (22). The enormous network of air defence centres across the U.S. that made up the Semi-Automatic Ground Environment (SAGE) largely subscribed to this topology, rendering it vulnerable to a set of carefully-targeted attacks. Baran writes that the "circular error probable" (CEP) of ballistic missiles--the CEP being the probability that a missile will land within a certain radius of its target more than 50% of the time--was sufficiently high that some serious consideration had to be given to ways of countering this.
This interest in designed more resilient network systems was not based on any will to increase the power of the United States over the Soviet Union in a hotly contested arms race. For Baran, the institutional structure of command was synonymous with the technical capabilities through which commands can be issued. As such, his central proposition was that a distributed command and control system, adopting the topology of a "non-hierarchical network", would be very difficult to destroy as no single node acts as a bottleneck. Applying this to SAGE would effectively make launching an attack on the United States a pointless exercise: the network would survive, allowing for operational command and control to persist, all but guaranteeing a full-scale counterattack. Baran writes:
The amount of redundancy necessary is that degree that causes attacks directed against the command structure to be no more effective than those directed against any other targets of the system. [...] With the redundancy of command function and command posts, possible great advantage may be made of secrecy. That is, the defence may play the old shell game; this time with a larger number of shells and a sizeable number of peas. The enemy must find all the peas in order to win. (13)
Baran's design was further elaborated in an 11-volume series of memoranda published in 1964. His proposals were not directly progressed, but he did succeed in having them declassified and published so they could be accessed by other American--and crucially, Soviet--scientists. Baran understood that if both superpowers had redundant, distributed command and control systems, a nuclear first-strike would be pointless, thus forcing a stalemate and effectively ending the viability and advantage of the surprise attack. This pacifying arms-parity induced by distributed networks was not to be the legacy of this document--rather, it was to be the theoretical prototype of packet-switching as a technique of redundantly routing information across a non-hierarchical network.