The Internet Then vs. Now: Timesharing to Routing

Bob Taylor & Larry Roberts were imagining a future Internet where computers could be gateways to each other in a distributed network. This was opposed to centralized AT&T lines. They imagined they needed to supplement the flow of information in the event a line making connection breaks down, similar to how neural networks operate in humans. This is the same today. They imagined users would have a personal computer, separated and freed from multiplexed computer terminals, yet still able to share their resources to be able to create a pool of information at a database. A challenge to surmount was that the operating system languages of different computers were becoming increasingly different.  At the time, most computer scientists at universities didn’t want to share their multiplexed computer’s resources. Terminals were rented by the processing time and the resulting data count. Costs could be expensive. In the time-sharing system, many users at multiple terminals were using the resources of one multiplexed computer performing batch processing, as opposed to accessing databases on disks or drums. This was at first how Taylor & Roberts had to approach their proposal at DARPA, imagining how they could find a way inside each of the computers at the time to connect disparate types of computers to each other, to share information. This is very different from how our network is today. Another challenge to surmount was the rate of data transmission. Instead of waiting for requests to be completed, like standing in a long line, Donald Davis and Paul Baran, who informed Roberts and Taylor of this, imagined data being packaged and delivered in the most time-efficient way possible, with computers estimating and informing each other on the fastest hops to take in a distributed network of computers together, so that information can be transmitted from A to B and from C to D, and so on. Although packet-routing was to stay, this vision of using resources inside the computer had to change, and with the help of Wes Clark, Donald Davies, and Frank Heart, it did for the ARPA-net and today’s Internet. They began imagining a network of microcomputers that could enable communication between greatly differentiated types of computer hosts. This was thought of as a subnetwork that could ensure the microcomputers were using separate resources from the hosts, and all speaking the same language.  The Interface Message Processor (IMP) became one of the first generation of packet-switching routers to run on its own resources connected between hosts via a special bit serial interface. Using its own language designed by the manufacturer to transfer bits, IMPs began acting as gateways, presenting information at the application layer to the host computers. An astounding feat! They probably did not imagine we would be utilizing their ideas in the next generations to come to build wireless IEEE 802.11 networks, interface with virtualized servers (live in the cloud) and see their concepts become widely accepted and used by the telephone companies.