Time-Sharing

Before time-sharing, using a computer meant batch processing: you wrote your program, punched it onto cards, handed the deck to an operator, and came back hours or a day later for a printout. If you had made a typo, you found out tomorrow. The machine was the scarce thing and your time was the cheap thing, and the whole workflow was arranged around that fact.

Time-sharing inverted it. One processor switches between many users fast enough that each feels as if the machine is theirs. Nothing about the hardware became conversational; the illusion of attention is manufactured by scheduling. But that illusion is the entire precondition for interactive computing — you cannot design a mouse, or a live document, or an undo command, for a machine that answers you tomorrow.

Why it mattered to this story

Engelbart’s NLS is unthinkable without it. NLS ran on an SDS 940 under the Berkeley Timesharing System, supporting up to sixteen workstations at once. The 1968 demo showed two people editing the same document from different cities, which requires that both of them be inside the machine simultaneously. That is time-sharing doing the work, invisibly, underneath the part everyone remembers.

The paradigm also produced the first real evidence that people would use computers socially. Shared file systems and messaging appeared on time-shared machines almost immediately and without anyone planning them, which is a pattern worth noticing: give people a common substrate and they will start talking on it.

How it ended, and how it came back

By the early 1970s the complaint was contention. With enough users, response time degraded precisely when the work was most interactive. Butler Lampson’s December 1972 "Why Alto?" memo at Xerox PARC made the argument directly: stop dividing one machine among many people and give each person a machine. The Alto — and eventually the Macintosh — is what that answer looked like.

The idea did not die; it moved. Renting slices of a remote machine you do not own, billed by use, is the operating premise of cloud computing. The economics that made time-sharing sensible in 1966 made it sensible again once the network, rather than the processor, became the scarce resource.