Networking: Vision and Packet Switching 1959 - 1968
Intergalactic Vision to Arpanet
2.3 Circuit Switching
In January 1878, the first telephone switch went into operation in New Haven Connecticut. Switching technology had advanced drastically over the intervening decades, yet the basic function had remained the same: interconnect users of telephones by creating circuits between them. Every telephone has a line, or circuit, that connects physically to a telephone switch. In the simple case of both the person making the call and the person being called are connected to the same switch, the caller dials the number of the desired person, the switch checks to see if the line is available, and if it is, the two lines are interconnected by the switch. The connection is maintained until one person hangs up his or her telephone, at which time the switch terminates the connection, freeing both lines for other calls.
Three characteristics of this type of switching, called “circuit switching,” are important. First, before the two parties can talk the circuit between them has to be created, and it takes time for a switch to check if a connection can be made and then to make the connection. Second, when a connection has been made, it creates a dedicated connection. No other party can reach either party of a dedicated connection until that connection has ended. Three, since switches are very expensive one accounting policy telephone companies implemented to recover their investment was to institute a minimum charge for every telephone call, generally three minutes. For voice calls that lasted many minutes, a minimum charge did not represent a problem. But communications between computers often last less than seconds, much less minutes. It was difficult to image how circuit switching could work efficiently for computer communications when such a system took minutes to make a connection, created dedicated connections so only one person, or party, could be in connection with another party, and had a prohibitive cost structure.
Although these issues were generally understood before the experiments of Roberts and Marill in 1965, they were once again strongly confirmed. The experiments also made it abundantly clear that the problems confronting computer communications were not only with the circuit-switching architecture of the telephone system. Host operating system software of the day assumed there was only one Host and all connecting devices were as if “slaves.” Hosts were not designed to recognize or interact with peer-level computers; the concept of peer-level computing did not yet exist. Thus, in interconnecting two computers, one had to be master and one slave. The problem only became worse if more than two computers wanted to interconnect and communicate. Nevertheless, the problem of Host software was considered to be solvable if a suitable communication system could be designed and made to work.
Fortunately, an inquisitive innovative scientist, Paul Baran, had already explored the problems of circuit switching beginning in 1959. By 1962, he had made his concept of a message-based communication system publicly known. Independently, in 1965, an English scientist, Donald Davies reached the same conclusions as had Baran and would coin its name: packet switching.