He traced some of the history of computing. First you have dumb stateless terminals (e.g. the IBM 3270) that connect to a central mainframe via half duplex connections, displaying forms but submitting them back to the server and relying on it to do all the processing. Then this evolved into terminals that are able to do full duplex real-time interaction with the mainframe, maybe a character at a time, but still relying on it to perform computations. Finally, “smart” terminals evolved that were able to store state and perform some computations on the client side.

Then, in the early 90s, the next step in the evolution of computing arrived: the browser, a dumb stateless interface connected half-duplex to a central server, capable of displaying forms but submitting the results back to the server and relying on it to do all the processing. Then you have ajax, a full-duplex communication in which the client remains dumb and stateless but at least a little more interactive. And so forth.

In his words, “There is no practical difference between the [web] browser and the [IBM] 3270, except that the browser can display better porn.”

Interesting to think that maybe all we’ve been doing so far with internet applications has been to go around in circles and reinvent the 70s.