A technique by which communication resources are allocated dynamically to multiple communicating entities. Messages between entities are partitioned into segments with a fixed maximum size. The segments, or packets, are passed through a store and forward switching network until they reach their destination (or are discovered to be undeliverable). The packets are reassembled, if necessary, into complete messages when they reach their destination. Packet switching, as it applies to electronic communication, was first proven feasible by the development of the ARPANET in 1969.

A packet switching network may provide a variety of levels of service, depending upon the sophistication of the underlying communication technology and the requirements of the network’s customers. The simplest packet switching networks provide only datagram service, which is unordered unreliable delivery of packets. Other networks may provide only reliable individually flow-controlled virtual circuits. The decision between datagrams, virtual circuits, or other modes of operation can be made independently for the internal operation of a packet switching network, and the interface that it presents to its customers. See also cell, frame.