[fitsbits] leap second alert

[Mark Calabretta]

I'm surprised at the apparent willingness of the astronomical community
(or perhaps just the FITS community) to accept a change in the
fundamental definition of UTC.

For a start, it seems to me that although the astronomical community may
be responsibile for the implementation of civil timekeeping it is not
their prerogative to change the definition of the system in general use.

Surely civil timekeeping is a matter of government regulation.  It is
perhaps nothing short of a miracle that all countries (are there any
exceptions?) agree to use a consistent timekeeping system, UTC.  There
would need to be general agreement between international governments on
any changes made to its definition.  

The basic definition of UTC ties it closely to the Earth's rate of
rotation.  Surely "Ephemeris Time" (the precursor of TDT/TT) was available
if a "uniform" timescale had been desired for international use.

Certainly TAI is available now.

When I originally responded to Demetrios Matsakis' call for opinions on
the matter I pointed out that it is often convenient that computer system
time (synchronized to UTC via NTP) is a close approximation to UT1.  In
applications where millisecond accuracy is not critical, LAST (local
apparent sidereal time) can be computed to within a second without having
to resort to table-lookups.  The particular application I had in mind
involved a graphical, time-changing, representation of the sky used as a
user interface in real-time telescope control.  Sub-second accuracy is
not required but errors of tens of seconds would be noticable here.
There must be a lot of other astronomical software which falls into this
category.

On questioning, Demetrios said that one of the protagonists for the
change was the NTP community itself since it "...takes a while to
propagate leap secs thru the hierarchy. Leap seconds are a tremendous
headache in the NTP world because they cannot be predicted."

While I sympathise with their difficulties (I've been there myself), it
is invariably true that IERS gives *plenty* of notice (several months
at least) of an upcoming leap second.  I can only conclude that NTP was
not designed with leap seconds in mind.  This would have to rank as a
Y2K-type flaw if you consider that leap seconds have been around for
nearly thirty years now.

In any case, my feeling is that anti-leap-secondists should be lobbying
for civil time to be changed from UTC to TAI (+ offset?).  The proposal
to change the definition of UTC appears to me to be a back-door route to
achieving this.

But if, in the end, civil time was to be converted to some uniform time
scale then why not go the whole hog and dispense with timezones as well.
Then at least civil time would correspond with solar time along one of
the Earth's meridians.  Move over Greenwich!

Mark Calabretta
ATNF