Networking: Vision and Packet Switching 1959 - 1968
Intergalactic Vision to Arpanet
2.6 Packet Switching
Packet switching represented a paradigm shift in communications technology. A paradigm shift that in combination with the digital revolution would, and has continued to, transform not only the nature of computer communications but communications of every kind, including voice. But few could have imagined such implications in 1966. There had to be working proof that a communication system could be architected based on a technology different than creating circuits; that in some as of then undiscovered manner, messages or packets might simply be launched into a communication network and be delivered error-free to their destination. That first working proof on a scale that could not be dismissed was the Arpanet, a story to be told shortly.
Baran and Davies both had training and experience that enabled them to intuit the implications of digital transmission for “intermittent” or “short messages.” The coming digital transmission circuits would be so much faster than the then existing analog circuits that the carrying capacity, or bandwidth, of the digital circuits would accommodate many users or messages at the same time. Making those implications even more dramatic was the burst-like and asynchronous nature of computer communications. Users frequently transmitted short messages in request for information, or in response to requests, with responses often being much longer messages. Furthermore, humans take time to respond, whether in thought or physical dexterity, whereas computers perform as if instant in comparison. These differences between computer and voice communications were not new to the computer scientists drawn to revolutionizing computer communications, for many of them had intimate knowledge designing and coding operating systems for computers, systems that communicated by passing messages. So what had been new to Baran and Davies had become common knowledge to those who would design and construct the Arpanet.
Packet switching as in the sending and receiving of packets might have been an obvious alternative to circuit switching, but exactly how to build such a system remained unclear. Hundreds of questions and concerns remained, such as how would the messages between computers be parsed into the presumably fixed length packets that would circulate within the communication network? Then how would the packets be reassembled into the original messages, especially if the packets arrived out of order? How would packets be buffered in the store-and-forward switches if the transmission possibilities were unavailable and the switches became overloaded? What about congestion, or error-correction, and so on?
Even the purity of packet switching came into question for there were compelling reasons for creating circuits between communication parties. Was there a combination of the use of circuits and of packets that might be optimal? Such as users, or programs, interacting with the system as if circuits were being created and yet the communication network functioning as if simply passing packets? This concept, that of virtual circuits, not real or physical circuits, but virtual circuits being created to better effect communications would become important albeit not at the time.
The solution of how to build large, multiple networks of computers will not be fully demonstrated until 1988 and, even then, the solutions that would be thought to dominate would with time change. But the world of future computer communications, future as in looking forward from 1965, would never have become what it will without the profound impact of packet switching.