[LEAPSECS] Caveat emptor!

[Rob Seaman]

Warner Losh wrote:


> Rob Seaman wrote:

> 

>> Universal time is time-of-day.  The current definition of UTC permits it to be used to recover an interval timescale.  Timestamps supply information about both time-of-day and an interval since an epoch.  More to the point the interval may be in evenly-tempered units or in jerked-about-by-the-moon units.  It may very well be of interest to different stakeholders whether the stamped event occurred at "noon" or rather whether it occurred so many thousands or millions of SI seconds before or after some other event.  The current definition of UTC preserves access to both.

> 

> It is actually the elapsed time in SI seconds since the fictions midnight, which differs from both solar midnight and mean solar midnight by some amount. This differs a little from the elapsed time in mean solar seconds as well.  Since it is expressed in SI seconds, which do not represent 1/86400th of the earth's rotation, they are a pure elapsed time that happens to be approximately the same as a time of day measurement.  The current definition of UTC doesn't answer the question of if it is an elapsed time or a time of day, since the difference between these two is small due to the addition of leap seconds.

> 

> So while UTC is steered to be a useful approximation of earth orientation, that does not make it a measure of earth orientation.


However you like it, although I don't perceive the significance here of drawing a distinction between the terms "approximation" and "measure", and don't think I used either word myself.

We often go exploring in thickets of complexity on this list.  For civil timekeeping, rather, the situation is quite simple, really.  The "day" in time-of-day is the synodic day.  The synodic day is the sidereal day (the natural rotation period of the Earth relative to the "fixed" stars) adjusted to compensate for lapping the Sun once per year.  Another term for this is mean solar time.  The technical elaboration of these and related terms can be as detailed as one likes and can pull in orbital mechanics, perturbation theory, general relativity and all sorts of arcane physics as well as maddening international standards deliberations, but the *meaning* of time-of-day remains clear.

By all mean let's discuss improving current civil timekeeping policies.  This then turns into a very traditional exercise in system engineering best practices.  The ITU has not engaged in such an exercise.  Due diligence is entirely lacking.  I happen to disagree with an assertion that redefining time-of-day is a good idea (or even possible in the way they mean), but there are normal engineering practices for performing trade-off studies and risk assessments.  Most fundamentally it is the attempt to redefine the problem out of existence that I disagree with.  Civil timekeeping (of course) is based on time-of-day.  Time-of-day (of course) is based on synodic timekeeping.  Fiddle around the edges, but the central notions have to kept straight if the edges are to be located in the first place.

Perhaps we might adjust the allowable errors (what distinguishes an approximation from a measure, I guess) to permit a greater range of excursion from the definition of time-of-day, but we can't just "forget the whole thing".  If the SI-second differed by even 1/1000 of 1% from 1/86,400 of a (whimsically evolving) synodic day, the entire scheme would come tumbling down.  We're lucky to have some flexibility, since the 1820 effective epoch of the SI-second remains "pretty close" to the current value underlying time-of-day.  But there are indeed two clocks and civil timekeeping depends on both.

The ITU-R is abrogating to itself the authority to redefine time-of-day.  The implications of such a change stretch far beyond the limits of radio-communications (as do the frequent discussions here of computing issues).  With authority comes responsibility.  The continuing lack of consensus here clearly indicates that the ITU-R has not satisfied their responsibility for any change to civil timekeeping.

So we're to "forget" the 2003 UTC Colloquium in Torino.  Fine.  Organize another Colloquium reaching out to a broader range of stakeholders.  More minds may result in a more flexible consensus and more creative ideas.

So we're to edit phrasing like "GMT may be regarded as the general equivalent of UT." out of what was previously CCIR 460-4.  Not so fine.  The collateral damage here isn't limited just to UTC, but rather will extend to the entire Universal Time family of timescales.  Universal Time has meant (an approximation to or measure of) Greenwich Mean Time for many decades.  Confusing that meaning is a very shortsighted exercise - a truly bad idea.

The ITU appears to be only willing to consider zero effort (or less than zero) "solutions".  By contrast, the astronomers here have repeatedly demonstrated the willingness to consider a wide range of possibilities - not surprisingly, some of these require actual engineering to accomplish.  If the ITU wants an easy way out, they can 1) call the new timescale something other than UTC, since it indeed will not be a flavor of Universal Time, or 2) pursue Warner Losh's proposal to significantly lengthen the leap second scheduling horizon by loosening the DUT1 bounds in prudently stepped and vetted intervals.

"Caveat emptor!"  "Oh, you're Celtic!"

Rob