[LEAPSECS] alternative to smearing

[Zefram]

Preben Norager wrote:
>Julian days count solar days.

They count whatever days they are applied to.  Originally that was solar
days, of course.

>julian days somehow count from solar noon, and that is why I wrote the
>julian period count apparent solar days.

That's a misunderstanding.  The Julian Date counts from noon of, again,
whatever days the system is applied to.  Just as the Modified Julian
Date counts from midnight.  The noon and midnight are not necessarily,
or even usually, the noon and midnight of apparent solar time.  They can
equally well be the noon and midnight of mean solar time, and indeed this
usage with various flavours of UT is probably the most common application
of JD and MJD.  When these day count systems are applied to non-solar
time scales, the noon and midnight are then nominal constructs that have
nothing to do with the sun.  It works just as well in all cases.

>                                         Without leap seconds in the civil
>time, then in a very long time-span the number of calendar days not
>nessesarily match the number of solar days.

This is true, and it is reasonable to argue as a result that when using
a non-solar time scale we should not use day labels that were originally
meant for solar days.  In principle, to apply JD or the Gregorian calendar
to TAI is as nonsensical as it would be to apply them to sidereal time,
or to Airy Mean Time (mean solar time on the prime meridian of Mars).
But in practice that decision was made around 1950 with the development
of Ephemeris Time, and baked into horological culture by the subsequent
decades of use.

The hypothetical TI that would succeed UTC as the basis of civil time
occupies a strange status here, which would look a great deal stranger if
we had not had this historical accident with ET.  TI would be essentially
an atomic time scale, with the same nature as TAI, and so by nature one
would think it nonsensical to use solar day labelling with TI.  But for
it to be used as the basis of civil time, from which local timezones are
derived which have a practical need to track the sun, it must necessarily
have a defined correspondence to solar day labelling.  The labelling
would be justified not by the time scale actually tracking solar time,
but by the intent that it be used to derive local time scales tracking
solar time.  This is a relatively unexplored view on one of the familiar
arguments against making such a change to civil time.

>                                            So in a very long time-span the
>number of julian days can no longer be used without cultural bias.

What do you mean by "cultural bias" here?  Are you just suggesting that
TAI, by its artificiality, is an example of cultural bias, whereas UT
is natural and independent of culture?

>                                                                   And when
>julian days can no longer be used without cultural bias the scientific
>community will have to choose between julian days and the gregorian
>calendar.

This doesn't follow at all.  The problems, such as they are, in applying
solar day labels to TAI's nominal days happen equally for all types
of solar day label.  Julian Dates and the Gregorian calendar were both
originally used for solar days, and are now both used for non-solar time
scales as well.  They are pretty interchangeable for these purposes.
What relevant distinction do you see between them?

>                                                       I think the only
>choice will be the gregorian calendar, because that calendar is the only
>one in universal use.

It certainly has a dominant position, used by most cultures worldwide.
The network effect argues strongly in favour of it being used for civil
purposes everywhere.  But that doesn't extinguish choice.

Your argument here for the civil and scientific use of the Gregorian
calendar doesn't seem to have anything to do with the choice of underlying
time scale.  The Gregorian calendar is dominant *now* in civil use.
It tends to be used for scientific purposes too, except in situations
such as those seen in astronomy where a linear count of days (such as JD)
is more convenient for computation.  How would a transition from UTC to
TI make any difference to these pressures?  How would it even require
a reconsideration of day labelling choices already made?

-zefram