Practicing for the New Millennium
Silicon Valley, 31 December 1998

by Brian Reid

Anglicans Online is written and published by people who have daytime jobs doing other things. As AO's publisher, my day job is that I am a computer network R&D manager. I am getting ready to write a long essay about Anglicanism and the Y2K situation, but today I was diverted by a different sort of problem, one year before Y2K strikes Greenwich.

The world's hard-core computer network gurus had a small crisis this afternoon at 4pm California time, but we handled it like the professionals that we are. We had a leap second, and we kept our composure.

There is one more second than there used to be. The memo announcing this, incredibly and amusingly dry, can be found in Time Service Announcement Series 14, where it says:

1. The International Earth Rotation Service (IERS) has announced the introduction of a time step to occur at the end of December, 1998.

2. Coordinated Universal Time (UTC) will be retarded by 1.0s so that the sequence of dates of the UTC markers will be:

1998 December 31 23h 59m 59s
1998 December 31 23h 59m 60s
1999 January 01 0h 0m 0s

3. The difference between UTC and International Atomic Time (TAI) is:

from 1997   01 Jul, UTC to 1999 01 January, UTC:    TAI-UTC= +31s
from 1999   01 Jan, UTC until further notice:       TAI-UTC= +32s

You have to love the signature at the end of the memo:

DENNIS D. McCARTHY
Director
Directorate of Time

In the old days, people kept track of the time by having master clocks in the hands of the Navy. The Royal Greenwich Observatory and the US Naval Observatory are two examples. First the navies made them out of brass, but now every Navy buys them from the low bidders, who make them out of silicon and cesium and xenon. In these modern days, EVERYONE keeps track of the time by asking a computer, even the Navy. The time might be defined by the rotation of the earth, but it is counted and told by computers.

At some point the computers have to learn what time it is. People whose life or livelihood depends on an accurate clock will buy their own atomic clock from one of the speciality instrument companies, but for the rest of us there is a fairly low-profile network of master time computers that tell each other the time, using a protocol called XNTP 3. We network people call it RFC1305; consider it to be the tablets that David Mills brought down from the mountain.

But you don't really want to read engineering scripture. You just want to know what time it is.

Your computer found out what time it is by asking some other computer, which in turn found out by asking some other computer. Each computer's clock is identified as being of a certain "stratum". Because I am a fanatic I run a Stratum 3 clock at my house. It finds out what time it is by asking a Stratum 2 clock. The Stratum 2 clocks find out what time it is by asking Stratum 1 clocks. And yes, the Stratum 1 clocks find out what time it is by asking God, or at least directing the question in his His general direction.

Well, not exactly. There are 73 working Stratum-1 NTP clocks in the world. Four of those are in the laboratory where I work. On New Year's eve at 4pm California time (midnight GMT), nearly everybody in the world who runs a Stratum-1 clock was watching it more carefully than usual to make sure that it got the extra second, because, gasp, if it didn't, then hundreds of thousands of computers would have the time wrong by one second and we would be responsible. While it might not sound too awful to have time time wrong by one second, it can be ruinous for some computers. If time is off by one second then two files that are supposed to be identical and have the same creation time are suddenly not identical any more. And given the collection of lawyers and barristers who are preparing to file lawsuits for Y2K problems, one might worry that they would want to rehearse by filing lost leap-second lawsuits.

Well, guess what. Our clocks made the jump. Here's the trace, in raw network-engineer language:

Slightly before 16:00 PST:

time to next leap interrupt: 179 s
date of next leap interrupt: Thu, Dec 31 1998 16:00:03
calls to leap process: 193
leap more than month away: 138
leap less than month away: 30
leap less than day away: 1
leap in less than 2 hours: 24
leap happened: 0

Slightly after 16:00 PST:

time to next leap interrupt: 86327 s
date of next leap interrupt: Fri, Jan 1 1999 16:00:04
calls to leap process: 194
leap more than month away: 138
leap less than month away: 30
leap less than day away: 1
leap in less than 2 hours: 24
leap happened: 1

Now this is actually very scary, because it says that the Earth is slowing down so much that they're going to do this again next year. A leap second two years in a row. Be still, my throbbing heart. It's going to be a busy year at the International Earth Rotation Service. Since 1972, which is presumably when atomic clocks were made accurate enough that navies could detect the need for a leap second, there have been 22 leap seconds.

If you are a network time connoisseur you will understand that the meat of the U.S. Navy's announcement is that the displays on the front of the Stratum 1 clocks were supposed to do this:

23:59:58, 23:59:59, 23:59:60, 00:00:00

But they didn't. We watched carefully. The darn clocks were aware of their noblesse oblige in being master clocks and just could not bring themvelves to display 23:59:60 as the answer to "what time is it?" So instead our master clock just slowed down a little bit between 23:59:58 and 00:00:00 so that it could tick 3 seconds in the space normally reserved for 2. At least that's what the front panel did. In my line of work watching an atomic clock slow down for a second or two is like watching a national politician pick her nose on camera during an important speech. People gasped "I can't believe what I just saw."

Then we all went home.

Happy New Year.

Brian Reid
Publisher, Anglicans Online