[LEAPSECS] Reliability

[Adi Stav]

On Mon, Jan 05, 2009 at 09:39:28AM -0700, Rob Seaman wrote:

>

> Lower limits are hard to pin down.  Human tolerance on a particular day 

> is not the same thing as the tolerance over a year or a lifetime.   

> Straining a tolerance for one human is not the same as straining it for 6 

> billion.  Human tolerances in general need to be interpreted in terms of 

> our infrastructure, not just personal perception as we walk from parking 

> lot to office.


How is that? That is to say, what problems could exceeding the
tolerance(s) cause? (Especially problems that time zones far from their 
reference meridians, DST switches twice a year, and the difference between 
mean and apparent solar time don't already cause). I'm not arguing that 
there arent's such problems, but I don't know what they are.

Only thing I can imagine that is not covered by time zones etc. is the 
minute of drift that will be experienced over a person's (or a system's) 
life time.


> The upper limit has been specified as a "statement against penal  

> interest" by the ITU.  Public enemy  number one of leap seconds says an 

> hour is the upper limit :-)


An hour makes a lot of sense from a usability point of view, because it
is the primary division of a day.


>>> Embargoing leap seconds (or their equivalent) for periods of decades 

>>> or

>>> centuries is the same as not making intercalary adjustments at all.

>>

>> Why is that? Even the Gregorian reform does not come into effect  

>> except

>> every one or two centuries. Yet it is followed exactly.

>

> Gregory revised the Julian calendar.  The fundamental standard remains  

> rooted in what the ancients discovered.  The proper comparison is to the 

> every four year scheduling of leap day opportunities - sometimes those 

> opportunities remain nulled out, but they still exist.


The Gregorian reform revised the Julian calendar, but it still had to be
introduced. It set a schedule that does not come into effect except once
every one or two centuries, which was followed to the letter. I think
this is impressive, even if it used a pre-existing mechanism.

A good parallel would be adding leap hours and using the existing DST
mechanism (not that I can't see other issues with it).

But here's a thing -- maybe suggestions for making leaps happen very
infrequently are seen as dishonest, "let it slide" in disguise. But it
doesn't have to be this way. I can see good, honest, technical reasons
for leaping every few centuries. You can use leaping mechanisms
that are simply not available when you have to leap every year or two. 
(For example, you can introduce a new time scale (UTC-n) every few 
centuries and deprecate the old one over decades as users switch to 
the new one. I can think of several technical advantages of such a 
system over leaping an existing time scale.) Another example -- the
Julian calendar did slide over a very long time, yet it did not stop 
people from fixing it, and that was even as its original definition 
did not prescribe the fix. If a definition of a time scale does
explicitly require a leap minute or hour in a very long time, why 
assume in advance that it will not be followed?


> The seasonal or diurnal trends in the calendar or clock need to be  

> sampled frequently enough to avoid significant quantization errors.   

> Leap seconds are productive from this point of view precisely because  

> civilians can ignore them.


I'm sorry, I don't understand :)


>> By the way, it can be argued that the smoothness property is not  

>> strictly

>> necessary for calendars. Consider popular and long-used artihmetic

>> lunisolar calendars, such as the Hebrew, Hindu, and Chinese calendars,

>> that intercalate their years to a resolution of a month.

>

> A very interesting observation.  What calendars does the world really  

> depend on for various purposes?  That is, what is the market penetration 

> of the Gregorian/Julian calendar?  I would guess nearly 100% in Europe 

> and North America.  What about the rest of the world?


I think they are used in conjuction with the Gregorian for different 
purposes, such as for public holidays of traditional origin. (As an 
anecdote, DST in Israel starts according to the Hebrew calendar but ends 
according to the Gregorian.) But I don't think it matters much for our 
purpose, because any culture which is in contact with the current global
civilization will be under a lot of pressure to adopt the Gregorian
calendar for many purposes, even if its traditional calendar is actually
superior for its needs. Prior to contact with the Western civilization,
the old Chinese calendar was used in China for a long time. I don't
think we have any reason to assume that the Gregorian calendar is 
superior to a lunisolar calendar and that those lunisolar cultures 
changed calendar for this reason rather than for reasons of culture and
trade influence.