A Brief History of the Internet's Origins - The Challenges That Motivated The Arpanet

The director of the Advanced Research Projects Agency in the Defense Department decided in 1966 to create a research network connecting the big main frames at US universities. By this point the first book at packet switching had already been published. Furthermore, it was quite clear that the universities would benefit from directly sending data from one mainframe to another without sending a tape disk via the postal service. The first computer network was the project's animating vision. Research in this context meant two things. The network itself would be a test bed for new networking ideas about how computers could directly talk to each other. Secondly, it was hoped that academics and scientists would demonstrate the Arpanet's value by using it to further their research.

It is hard to exaggerate how challenging computing was in the beginning. The ARPA director, Robert Taylor, had three computer terminals in his office linked via dedicated 56 kilobit lines to the three computer projects that the agency was supporting, namely the Genie project at Berkeley, the System Developments Corporation in Santa Monica, and the C.T.S.S project at MIT. What Taylor noticed was that local user communities formed around each mainframe because files and messages could be transferred from one local participant to another. Academic researchers in different fields began collaborating because they met each other via the mainframe. It was like a water hole for academic animals. So Taylor saw that the full potential of these mainframes would only be achieved if distance was no longer a barrier. In other words, a wide area data network was required. Moreover, it must be easy to access these mainframes via standardized, interoperable terminals. Taylor realized that access interoperability and wide area networks were the two keys to fully exploiting the power of computers. 

Robert Taylor went to the head of ARPA, Charlie Herzfeld, and pitched the idea. Herzfield liked it so much that he funded it in a few weeks by taking a million dollars out of ballistic missle defense research. Three years later the Arpanet was launched with four participating sites, UCLA, University of Santa Barbara, SRI, and the University of Utah. It took two more years to eliminate bugs from the system. The Arpanet could do three things once it had been fine tuned: file transfer, accessing remote computers and also printers. But the system grew extremely fast with NASA, MIT, and the Rand Corporation all joining by April 1971. At that point the system had 15 nodes and 23 host terminals. 

Robert Kahn wrote the first internetworking protocol, Network Control Protocol (NCP). This protocol did two things. First, it set up a connection between two computers. Secondly, it created an unidirectional, flow controlled data stream. Note there was no error detection or ability to correct data errors.