On the Addressing Problem for Loop Switching

The methods used to perform the switching functions of the Bell System have been developed under the fundamental assumption that the holding time of the completed call is long compared to the time to set up the call. It is thus sensible to hold portions of a route while the attempt is made to establish the connection. In considering certain forms of communication with and among computers, as well as the consideration of many schemes for time division switching, the possibility arises that a message, with its destination at its head, might thread its way through a communication network without awaiting the physical realization of a complete dedicated path before beginning on its journey. One such scheme has been proposed by J. R. Pierce, 1 and may be called "loop switching." We imagine subscribers, perhaps best thought of as computer terminals or other data generating devices, on one-way loops. If a meassage is destined for a subscriber on another loop it proceeds around the originating loop to a suitable switching point where it may choose to enter a different loop and continue the process until it reaches its destination. The question now comes up, how the message is to know which sequence of loops to follow. A sufficiently complicated memory in the originating loop might, of course, look up an appropriate route, and then attempt to seize a complete path; but this is the old and perhaps in-