[LEAPSECS] Cheating means more planning, not less

[M. Warner Losh]

In message: <5DE48B7A-0D30-4580-B110-7687A75A2904 at noao.edu>
            Rob Seaman <seaman at noao.edu> writes:

: Identifying the length of the civil day with the length of the mean  

: solar day is the key to providing that coherence.  (True now on Mars  

: as well as Earth.)  The mean solar day is just the sidereal day plus  

: the synodic correction for lapping the sun once a year.  The mean  

: solar day is a global phenomenon.  The eccentricity of the Earth's  

: orbit and the tilt of its axis (etc) add periodic terms that average  

: out.  Latitude and politics overlay local variations that are a  

: distraction from the central issue.  Tidal slowing, on the other hand,  

: represents a global long term secular trend.  A trend with global  

: implications demands a global solution.


Correct.  However, the die was cast on this in 1958 when the second
was defined in terms of atomic behavior.  At that point, the game was
up, since the basic unit of time was decoupled from the day.  We
transitioned from having rubber seconds, to having rubber days.  I
suppose we could push this back further when the second was defined in
terms of the mean solar day in 1900, since that changed a division of
a day, to the day being so many seconds.  A subtle difference that
appears to have been lost on the people taking this first step, at
least at the time.


: The trend just happens to be slow enough to permit cheating.   

: Consensus based planning is necessary *especially* if we decide to  

: cheat.  Cheating is ultimately fruitless over the long term, no matter  

: what.


Yes.  First, people cheated with rubber seconds (and why not, since
that's how people cheated before the fixing of the length of the
second based on atomic behavior: seconds were always rubber since they
were defined in terms of a day that varied in length).  Doing this
similar cheating with atomic clocks presented many operational
problems...

Second, people cheated with leap seconds.  This cheat has presented
many operational problems.


: The ITU has a responsibility to consider options with a long term  

: future.  A permanent embargo on leap seconds does not have one.   

: Whatever action the ITU takes, it should be fully and carefully  

: planned and not obligate our descendants to clean up an embarrassing  

: mess.


This is both true and false.  Assuming, for the sake of argument, that
we take the position that the ITU has to have a permanent solution to
the clock skew problem, that cannot involve leap seconds.  In time,
they will be needed more than twice a year, then more than 12x a year,
and finally, more than once a day.  UTC, as defined today, dies once
the length of day is 86401 seconds[*].  With these long-term problems on
the horizon, I agree that some solution is needed.

A leap second embargo is one radical idea "don't sync the clocks:
publish the delta."  The psychological aspects of this are nice, since
time won't drift more than a few minutes in anybody's lifetime for
several dozen generations yet.  This is as long term as other
'permanent' solutions the ITU has promulgated.  So I'm not sure I
understand this criticism.


: >> Only one - standard time based on mean solar time - has ever been  

: >> shown to be *practically* workable.

: >

: > Two: standard time plus daylight saving time is the other

: 

: DST is a trivial gimmick layered on standard time.  Standard time is a  

: global system layered on the mean solar day.


But UTC isn't layered on top of the mean solar day.  It is merely
synchronized to the mean solar day.  It is based on the atomic
second.  And many legal times are being transitioned to UTC.

There is a subtle difference between your statement and mine, but an
important one.


: Ideally we will come out of this exercise with an improvement to  

: standard timekeeping.  Wouldn't it be more fun to pursue that project  

: rather than playing an endless game of whack-a-mole with ITU politics?


Warner

[*] Well, it would fail when the day was ~86400.033s since that's when
we'd cross the once a month threshold for when the leap seconds can
happen.  When it crosses 86401, though, we're no longer able to use
the same notation we're using today with 23:59:60 although I suppose
it could be extended to any hour, then any minute, etc.  At some point
it becomes totally unworkable, and I arbitrarily selected "once a day"
although my hunch is that it is somewhat before then.