From Network to Patchwork 95
Department ordered ARPA to design a “survivable” network that would last
long enough in a nuclear strike to send a “go-code” to guarantee “second-
strike capability.” “There was a clear but not formally stated understand-
ing,” noted Baran, “that a survivable communications network is needed to
stop, as well as to help avoid, a war.”^38 A network that can survive an enemy
attack could ensure the threat of the mutual nuclear annihilation—a threat
so cataclysmic that it would rationally deter (Baran and his military supe-
riors hoped) either the Soviets, the Americans, or any other nuclear power
from striking first.^39
Baran’s inspiration for packet switching as a way to build a survivable
network traces back to Warren McCulloch’s cybernetic conception of the
human brain as a complex and resilient logical processor. As Baran reported
in an interview with Stewart Brand, “McCulloch in particular inspired me.
He described how he could excise a part of the brain, and the function
in that part would move over to another part.”^40 The same interview lists
McCulloch and Pitt’s 1943 paper on neural networks as a sensible refer-
ence, although Baran also noted that he was reading more broadly in the
“subject of neural nets,” a literature that probably included McCulloch,
(a) (b) (c)StationLinkFigure 3.2
Three network types: (a) Centralized, (b) decentralized, and (c) distributed. Source:
From Paul Baran, “Introduction to Distributed Communication Networks.” On Dis-
tributed Communications, RAND Corporation Memorandum RM-3420-PR, August
1964, 2. Reproduced with permission of The Rand Corp.