TimekeeperDavid Mills

(Reading Time: 4 minutes)

Born in California, Mills was diagnosed with having glaucoma since birth. A surgeon managed to save partial vision in one eye, but his vision was bad enough that he attended a school for the visually impaired. His 11th grade teacher told him, “You’re never going to get to college.” Mills said that was “like waving a flag in front of a bull.” When Mills got into computers, there was no such field as “Computer Science,” so his degree from the University of Michigan was in Electrical Engineering. Then he started grad school. “I had the second dissertation at Michigan in what was then called computer science, so it was quite new,” he said in a 2004 interview for the Charles Babbage Institute’s Center for the History of Information Technology. He earned his Ph.D in 1971. That was in the era of “big iron” (as Mills put it) — giant computers. “What started it was the IBM 360,” he said [see my profile of Fred Brooks for more on the System 360]. Mills taught at the University of Edinburgh, Scotland, before returning to the U.S. to teach at the University of Maryland, where he was denied tenure. He went to work in industry, which was lucky for us all. He took a job at COMSAT, a public, federally funded corporation intended to develop a commercial and international satellite communication system, which was one of the contracted organizations to the Defense Advanced Research Projects Agency, DARPA, which was working on a Big Idea.

So back to the “big iron,” which was beginning to rust. “The philosophy was, the computer cost so much that you had to have one of them for the university, and you couldn’t afford to decentralize, because that would strain dollars from the central facility. So, what happened with DARPA was that we began to believe that it wasn’t just big iron time-shared computers, it was decentralization. It started out with graphics displays. Obviously, centralized graphics displays were silly. So they moved them out into smart, relatively small computers with graphics displays on them.” To implement that, Mills created the “Fuzzball Router” — the first routers for the NSFnet, the National Science Foundation’s early 56kbps network. And that growing decentralization — that switch from stand-alone mainframe computers to interconnected smaller computers creating what became the ARPAnet, DARPA’s network that became the Internet — is what created a new problem: time synchronization. Those various computers really needed to all be set to the same time in order to properly share information with each other.

Mills, at COMSAT, went about solving that problem himself in 1977. The final result was the Network Time Protocol, which is still in use today. There are a number of synchronized Internet Time Servers which billions of computers and other Internet-connected devices connect to periodically to synchronize their clocks. NTP uses the intersection algorithm to select accurate time servers, and is designed to mitigate the effects of variable network latency. It can typically maintain time to within tens of milliseconds, and under ideal conditions can achieve better than one millisecond accuracy in local area networks. Since the “exact” time is open to interpretation (an averaging of multiple atomic clocks run by government agencies) and technological realities (the speed of light, for instance, since the farther away a time server is, the longer it takes for the synchronization signal to get back to the requesting device, creating an inaccuracy), Mills’ time protocol works by consensus, taking in all of these factors and more, to set the “correct” time. That meant, in Mills’ terms, figuring out which peer systems were “truechimers,” and which were “falsetickers” when it came to accurate time. His protocol figured that out. “I always thought that was sort of black magic,” said Vint Cerf, a pioneer of Internet infrastructure.

Dr. Mills at the University of Delaware in 2008. (CC 3.0 by Raul654 via Wikipedia Commons, cropped and enhanced.)

Mills was another of those pioneers of Internet infrastructure, serving as the chairman of the Gateway Algorithms and Data Structures Task Force, which later became the Internet Architecture Task Force, and wrote the first FTP (File Transfer Protocol) client software. He also served on the Internet Configuration Control Board. “That name is revealing,” he said. “We didn’t control anything. We couldn’t call it something that the government could construe as managing. We didn’t manage the Internet. We were just a committee down here that solves problems — configuration problems.” It later was renamed the Internet Activities Board. “What does that mean? “It doesn’t mean anything.” But the group, all brilliant designers, figured out various problems and solved them. “We were building standards, and we were doing so in a private way. ANSI and ITU [the American National Standards Institute, and ITU is the International Telecommunications Union] considered us irrelevant. So the idea was that, if we really wanted to standardize something, it wasn’t going to be with the international agencies, we had to do it ourselves. I think since then by the way a lot of progress has been made, to interface with ANSI and ITU. But back then we had to do it ourselves.” Good thing: it would have otherwise taken years for the problems to be solved. By 2004, he marveled at how commercial companies took things over and expanded the network to levels he could have never predicted. “I’ve got 100 megabits to [my] computer, anywhere I want to go,” he said. “It’s the flow of mice in a big pipe, instead of the roar of elephants down a tiny pipe. Quote me on that.”

Mills maintained the Network Time Protocol until the early 2000s. Mills also inspired the author of the Internet “ping” to write that utility (according to that author, Mike Muuss, Honorary Unsubscribe for 19 November 2000), and suggested an amusing “backronym” for it: Packet InterNet Groper. He also worked on version 4 of the Internet Protocol itself, which is still dominant today. Mills spent most of his career as a professor at the University of Delaware, which presumably saw fit to give him tenure. He eventually went completely blind and retired in 2008. Today he’s still best known for NTP, for which he has been dubbed the Internet’s “Father Time”. He received the 2013 IEEE Internet Award “for significant leadership and sustained contributions in the research, development, standardization, and deployment of quality time synchronization capabilities for the Internet.” Dr. David Lennox Mills died in Delaware on January 17, at 85.

From This is True for 21 January 2024