Apocalyptic Networks

Soon after, Paul Baran, a researcher at the RAND corporation, a Cold War think tank, formulated a key plan to realize a networked model of communication that could survive nuclear attack. Baran feared that the highly centralized model of communications used by both civilian and military telephone systems was vulnerable. One good hit on a city center would ensure that communications would be destroyed.


baran's networks


In its place, Baran developed a system of distributed communications in which each point functions as a node, the network’s common functions dispersed equally among the nodes. Designed not for present efficiency but for future survivability even after heavy damage during nuclear war, Baran’s system breaks messages down into discrete units or “packets” and routes them on redundant paths to their destinations. With expected transmission errors a fact of life in a post-Apocalyptic environment, his system allow damaged portions of a message”—instead of the whole thing”—to be resent. Following the model of urban decentralization, nodes would be located in the countryside, avoiding vulnerable city centers. In everyday operation, Baran’s “packet switching” system has the advantage of allowing individual sections of messages to be rerouted or even retransmitted when necessary and, as computers tend to communicate to each other in short bursts, takes advantage of slowdowns and gaps in communication to optimize the load on the lines.

Baran’s distributed network was meant to preserve hierarchy, not undo it. His goal was to maintain the centralized, top-down chain of command so that the other alternative”—giving individual commanders in the field authority over nuclear weapons”—would not be necessary. A victim of politics, Baran’s network was never built as he envisioned it, but his basic idea of the distributed network and packet switching would be incorporated into ARPANET, the first successful inter-city data communications network.

Established after the launch of Sputnik to recapture U. S. scientific superiority, the Department of Defense’s Advanced Research Projects Agency (DARPA) funded science and engineering programs in universities throughout the country, spurring on the development of high technology through projects such as the Orion nuclear powered rocket. To build community and overcome isolation between the fifteen offices of the Information Processing Techniques Office, the program funded ARPANET to link together researchers working with computers. ARPANET’s planners hoped that community would emerge through the experience of working together informally with shared resources. Developed to make remote time-sharing of resources work more smoothly, email soon became the network’s primary use.

Although ARPANET fostered a more distributed culture, encouraging informal, bottom-up management and interventions into the net, the model of distributed communications could not be fully implemented. ARPANET itself was distributed, but it was designed as an abstract layer, not as a separate physical network built from the ground up. Instead, it used existing long distance telephone lines leased from AT&T that, in accordance with the centralized model created a century earlier, joined switching stations in city cores. Had a nuclear war taken place, ARPANET would have been destroyed immediately.

Moreover, ARPANET emphasized the use of Interface Message Processors (IMPs), mini-computer interfaces allowing locally-based hosts to interface with the network. As there was generally only one IMP per city, the effect was that if ARPANET as a whole was distributed, at the local level it was highly centralized. Failure of an IMP not only meant that local hosts would fail to reach remote machines, it ensured they could not communicate with each other. By the mid-1970s, research-oriented networks such as ARPANET and the National Science Foundation’s NSFNet proliferated. Eventually these diverse networks would be linked by a single network of networks dubbed the Internet. NSFNet’s rapid growth during the 1970s made it the dominant entity in the early Internet. The NSF implemented communications between regional networks through a “backbone” leased on lines from AT&T and offered central hubs in each city to which local users would connect. The result was the final undoing of the distributed model.

With the privatization of the Internet in the 1990s, network topology has continued to centralize. Driven by profit, the need to build rapidly, and shackled by the difficulties of negotiations for new rights-of-way, telecommunications carriers follow existing systems of networking established by telephony. Interconnections are between major nodes located in city cores. Within cities, fiber optics can be laid down more inexpensively and higher capacity, short-distance networks can be built relatively easily. If, following AT&T’s breakup, there has been a proliferation of long-distance carriers carrying both data and voice traffic, these still access the local central office for distribution. High-speed private backbones operated by companies such as Level 3, Global Crossing, AT&T, or MCI now compete to offer connectivity around the world, but as data travels from point to point, it inevitably passes from network to network through a handful of peering points, interconnection sites between networks that are, again, located in city cores. This further concentrates the network, privileges the bigger players, and increases the divide between a digital hub in the city core and the digital desert beyond.