[LEAPSECS] Time-of-day

Rob Seaman seaman at noao.edu
Mon Mar 7 10:55:19 EST 2011


Poul-Henning Kamp wrote:


> Some languages, amongst them Danish, have two different words for "day" ("dag") and "24 hour from midnight to midnight" ("døgn"), but english notably does not.


Interesting! English does have "daylight" versus "a day", but your point is taken. (And there is the notion of astronomical twilight to add to your list, too.)

By all means define your terms. This is a good example of investing resources into first characterizing a problem before attempting to solve the problem.

My sincere apologies for my failures at communicating. Let's try again.

The issue is designing a system that uses clocks to mimic time-of-day. Time-of-day means time-of-synodic-day. It does not (typically) mean time-of-local-apparent-day-on-this-gregorian-calendar-date-of-the-year. It does not mean time-related-to-an-ensemble-of-atomic-clocks. It most certainly doesn't mean any-old-number-I-feel-like-announcing. As Poul-Henning says, it *could* be taken to mean all sorts of things, but most of these are unhelpful.

The "time-of-local-apparent-day-on-this-gregorian-calendar-date-of-the-year" meaning is unhelpful because it varies in both time and place.

The meaning, "time-related-to-an-ensemble-of-atomic-clocks", is unhelpful because it is circular. The whole point of the exercise is to design a clock system that mimics an external phenomena. A calendar mimics the Earth's annual revolution around the Sun. A clock mimics the Earth's daily rotation about its axis.

I'm sorry - the ITU simply does not have the mandate (defined in both Danish and English as "cojones") to redefine the length-of-day to be so far off as either 86399 SI-seconds or 86401 SI-seconds. What is that thing in the middle that they are constrained to bracket? That thing is the synodic-length-of-day of 86400+epsilon SI-seconds.

A modern clock ties time-of-day to the SI-second. (Or perhaps we can have a long digression now about the meaning of the words "clock" and "chronometer" :-) In order to have a coherent discussion we indeed need to settle on a common vocabulary for both "time-of-day" and "SI-second". I'm rather bemused that this seems to provoke such a strong response. I haven't been making any sort of case in the last few messages about *how closely* time-of-day needs to be simulated.

To solve a problem first focus on defining the problem, not on attempting to cut your perceived opponent's legs out from under him. Leap seconds are a means to an end. You may view them as *a* problem, but they are not *the* problem. The engineering problem is civil timekeeping. Leap seconds are part of the current engineering *solution* that some people want to change.

It is truly remarkable how messages recommending that a coherent system engineering process be followed - requirements discovery, trade-off studies, risk analyses, etc. - provoke such a strong response. At any rate, identifying a definition for time-of-day is part of the exercise of characterizing the common shared problem space, not of evaluating candidate solutions. Which is to say that this time-of-day discussion is at the opposite end of the entire process from any of the innumerable notional discussions we have had about leap-seconds.

What is "time-of-day" if not "time-of-synodic-day"? What is it that clocks are simulating?

I hope this helps clarify rather than muddy the waters...

Rob



More information about the LEAPSECS mailing list