One evening I sat down to sketch out a suitable network
and found it very simple in principle. Donald Watts Davies
In the early ‘60s concern arose that a misunderstanding between the two superpowers could prove disastrous. Both had nervous fingers on nuclear triggers. If command and control should fail there could be catastrophic results.
1960 RAND Corporation calculated that a stand-alone cable network with nationwide reach to all the US command centres and launch sites would need to be deeply buried if it was to withstand a nuclear attack. Building this network would cost over $2bn; Paul Baran at RAND tasked with seeking a low-cost solution.
He first considered the small AM radio stations criss-crossing the USA. Might they be annexed to relay commands or messages in a crisis? The military agreed that the president speaking to calm the nation could use this approach, but control needed more reliability in connection and bandwidth.
He decided against an analogue approach as telephone signals deteriorated across distance. His network needed to route around areas of damage and travel large distances without losing quality – he chose digital. He decided to chop a message into small chunks to be known as message blocks. These blocks could fly speedily around the network without overloading it.
Nodes could select the best route to a destination and route around damaged parts of the network. Each block would have a sequence code so on arrival it could be reconstituted correctly. An error-checking system would identify if the recipient had received corrupt or incomplete messages and request a re-send.
1964 Baran’s report ‘On Distributed Communications Networks’ defined his network approach. He and RAND decided not to patent or classify the thinking as its merit was in its ability to defuse any misunderstanding between the two superpowers; if it was plagiarised that might just be good news.
Quite independently, and with no knowledge of Baran’s work, the UK’s National Physical Laboratory came up with a similar solution.
Donald Watts Davies joined NPL in 1947 and set about emulating what was happening on USA campuses. He spent a year at MIT in the mid ‘50s and his mid ‘60s role took him to a conference in California where he visited Dartmouth to see the BASIC language and also MIT and RAND to learn about time-sharing systems.also MIT and RAND to learn about time-sharing systems.
Davies realised that long network messages blocked other users. He therefore applied the notion of time-sharing to the process, slicing up the network’s facilities between users. From there he realised that messages needed to be broken into pieces, sent in segments and reassembled. He called these segments ‘short messages’ but renamed them ‘packets’; it was Davies who originated the name ‘packet switching’.
1965 When Davies published his paper the Ministry of Defence advised that Baran had beaten him to it.
Roger Scantlebury, a member of Davies’ team, attended an Association for Computing Machinery meeting in Gatlinburg, Tennessee to present a paper. Also presenting was Larry Roberts of ARPA who was outlining early thinking on what was to become ARPANET.
It was in this roundabout way that ARPA became aware of packet switching; the organisation had not heard of Baran’s work directly. However it was very quick to act, hiring Baran himself as a consultant to develop packet switching as the basis for ARPANET.