[LEAPSECS] ISO TC 37

[Warner Losh]

On Jan 18, 2012, at 3:15 PM, Rob Seaman wrote:


> IWarner Losh wrote:

> 

>> Universal Time is an abstract definition.  It wasn't designed at all.  It models the time of day, on the average, of an important historical observatory in a nation that had the political clout to get its observatory named primary over all the other nations that had good observatories that wanted the honor.  Newcomb's equations of time are just this.

> 

> If the Jeopardy answer is:

> 

> 	"It models the time of day, on the average"

> 

> the question is:

> 

> 	"What is mean solar time?"

> 

> Steve's page elaborates on the historical aspects:  http://ucolick.org/~sla/leapsecs/timescales.html

> 

> In 1884 they may have debated different options for the Prime Meridian, but they were clear on the concept of the synodic day.


Universal time is an abstraction, as you say.  It models the synodic day.


>> Who has an actual requirement for an approximation of UT to 1s?

> 

> Almost everybody, but in this context many astronomical applications and systems.  And the "Draft Revision to ITU-R Recommendation TF.460-6" does not describe a time scale that approximates UT to any tolerance.  An unbounded perpetual secular drift is not an issue of precision.


A few people have a 1s error bar as a requirement.  More people may have a requirement that the error bar not be unbounded.  I strongly contest that almost everybody needs an error bar of 1s.


>> While I have no doubt that there will be issues, overstating them on time scales less than the current leap second rate and not providing an approximate DUT1 where problems occur does lessen the strength of the argument contained here.

> 

> 

> All such systems will require mitigation against this change.  The plan would certainly be to address issues before the standard takes effect, not wait for problems to occur.

> 

>> When the 2006 leap second happened, I happen to know of at least one company that spent about a million dollars to ensure that all their systems properly worked with the impending leap second.

> 

> And where is such anecdotal evidence collected and analyzed?  I don't discount it - and perhaps you can return the favor - but obviously the ITU process has not taken any such evidence properly into account if it isn't written down anywhere.

> 

> Leap seconds are a means to an end.  There are more nuanced positions than redefining UTC as TAI.


I'm not aware of any comprehensive study that shows what the costs would be on either side of the debate.  Some folks have given a study for their org, but that's it.


>> Bugs were found in the software (since untested software is buggy software), the GPS receivers, and some third party gear that had been integrated in to their systems.  We were a small company (sold a few years later for $12M), and I'm sure that we weren't the only company that experienced costs associated with that leap second.

> 

> These are costs, whatever else is true, that were accrued through not adhering to the standard in place at the time.  Astronomers and astronautical engineers have been using the standard as intended.

> 

>> Yet experience has shown that people just know that minutes always have 60s, and getting them to be pedantically correct continues to be a significant challenge.

> 

> They don't need to grok UTC to be correct.   They need to understand that time-of-day and atomic time are two different things.  And they will remain two different things if UTC is redefined.  Rather than increase the community knowledge of timekeeping issues, such an action by the ITU will only obscure it further.


They don't even understand that Rob.  And they do actually need to grok UTC to implement it correctly.  That hasn't happened yet after 40 years of the standard being in place.

Warner