The method of transmitting chopped-up ("packetized") messages across a computer network. Messages get broken up at the initiating computer; the packets are sent out like postcards onto the network with the address of the receiving computer, which reassembles the message upon receipt of the packets. The ARPAnet, Internet, and Ethernet are all examples of packet-switching networks.
Packet switching ensures that all messages are created equal; the network is neutral with respect to content. This neutrality has its drawbacks: Services that require rapid, high-bandwidth transmission, such as video, break down in a neutral packet network, unless it has very high capacity. But the advantages are greater: Because there are no inbuilt preferences, the network is flexible and extensible; any use of the network that can be imagined can be implemented, by anyone. Efforts are now being made to prioritize packets, which may have a chilling effect on innovation.
Donald Davies, a researcher at the National Physical Laboratory, began developing computer network communications in 1965. In his papers, he coined the term "packet switching". His proposals for a national packet network in Great Britain were dismissed by the institution in charge of communications, the General Post Office.
In On Distributed Communications (1964), Paul Baran worked out many of the difficulties in implementing packet switching in distributed networks, including how to handle lost packets, error checking, and addressing schemes.
Only at a 1967 Gatlinburg computer conference did Larry Roberts learn of the independent work done by Donald Davies and Paul Baran, which was incorporated into (and greatly influenced) his plans for ARPAnet.