In 1976, Mr. Cerf and his colleagues in the R.& D. office of the Department of Defense had to make a judgment call: how much network address space should they allocate to an experiment connecting computers in an advanced data network?
They debated the question for more than a year. Finally, with a deadline looming, Mr. Cerf decided on a number — 4.3 billion separate network addresses, each one representing a connected device — that seemed to provide more room to grow than his experiment would ever require, far more, in fact, than he could ever imagine needing. And so he was comfortable rejecting the even larger number of addresses that some on his team had argued for.
“It was 1977,” Mr. Cerf said, in an interview last week. “We thought we were doing an experiment.”