Everything in this is my guess at what might be happening next. There
is no reason to believe that any of it is accurate or will come to
pass. This is all just thinking out loud.
%%%%%%%%
The Torino colloquium posits that radio broadcasts of time signals
should principally provide a uniform atomic timescale to be called
something like TI instead of the current time-of-day (UT, earth
rotation).
This recognizes that
--> time-of-day (UT, earth rotation) is not uniform
--> the attempt to broadcast time signals using anything other than a
uniform scheme has historically been known to cause some form of grief
to some systems, even though to date this has not been demonstrated to
be a showstopper for any particular system
--> the need for uniform time interval in order to reduce the
complexity and likelihood of error in rarely tested sections of
software and hardware
And it is basically a statement that the need for systems to have
uniform time interval to be the principal signal has begun to outweigh
the need for humans to have time-of-day which has heretofore been the
primary signal.
%%%%%%%%
Nevertheless, I believe that the Torino colloquium asserted that it is
important that radio broadcasts of time signals continue to provide UT
to an accuracy sufficient for navigational needs of users now, and
also for human society and civilization now and into the foreseeable
future.
This value of UT is probably best provided as some form of difference
from TI. The difference (TI - UT) is expected to grow roughly
quadratically as centuries pass. This growth is due to the overall
tidal deceleration caused by the moon, sun, and other bodies in the
solar system. Nevertheless, there will be unpredictable fluctuations
in the rate of growth of the difference. The frequencies of these
fluctuations spans the spectrum from minutes to millennia due to
effects from weather, seasonal flows of surface water, annual to
decadal currents in the oceans, decadal and longer currents in the
outer core, and century to millenial climatic variations such as ice
ages coupled with crustal deformations caused by ice sheets.
%%%%%%%%
Implementing this change in broadcast time demands education.
>From the time that a plan is adopted, an educational program akin to that
used to introduce the metric system should be undertaken.
This means at the least curricula for children in school, and color
magazine inserts for the Sunday paper.
The educational program should make at least this much clear:
--> the rate of UT is not constant
--> simple differencing of values of UT does not produce an interval
in SI seconds, does not give a precise result, and is not something
they want to do for purposes other than social scheduling
--> in some 5000 years it should be considered perfectly normal that
the value of TI might be
7001-01-01T12:00:00, JD(TI) = 4278123, Thursday
when the value of UT might be
7000-12-31T12:00:00, JD(UT) = 4278122, Wednesday
because UT will have accumulated one full day less than atomic time
There needs to be a significant educational effort which gives examples
of how big the (TI - UT) difference is expected to be when, and has
exercises quizzing students about things that should and should not
be done using TI and using UT.
%%%%%%%%
Information systems should be modified, probably to represent all
timestamps as a data structure that contains TI and a difference
( TI - UT ). Isolated systems can simply keep time with their own
clock and set the ( TI - UT ) difference to NaN as a flag that there
is no information about earth rotation.
%%%%%%%%
Now to contemplate some technical characteristics of the changeover
from broadcasting UT to broadcasting TI.
We are told to expect that TI matches UT at the changeover, so
existing hardware keeps working with no perceptible glitch.
We are told to expect that the changeover will not happen any time
soon, and that it will not happen without significant warning.
This provides Agilent, Bancomm, Symmetricom, TrueTime and every
other vendor of precision timekeeping hardware with sufficient
lead time to develop new versions of their products that can
handle both TI and the provided difference with UT.
Systems using this new TI diverge from UT slowly, but the rate is slow
enough that, combined with the adequate lead time, everyone should
have ample opportunity to buy new hardware and install new software.
Everyone, everywhere should be warned that it will be necessary to
decide which time to use, but that their default choice means their
clocks drift away from the sun in a fashion that becomes an hour after
500 years. The "everyone, everywhere" part is tricky because it means
all owners of clocks, all legislatures, all systems maintainers, and
all authors of specifications that refer to values of time interval or
time-of-day. (This is going to force more scrutiny of this change
than the ITU or SRG would ever have wanted, and it will force them to
provide compelling, point-by-point justifications for the change.)
On many institutions and systems this forces choice of an offset of a
few seconds soon, or risk an offset of much larger magnitude centuries
hence. In most cases it will not be a huge problem if they decide to
buy new hardware and make the switch of some of their systems back to
UT relatively soon. There will just be a jump of a few seconds, which
is trivially absorbed by most systems that are currently in operation,
especially if those systems are accustomed to working in UT.
For those institutions which do not make the decision to buy new
hardware and switch back to UT, and who later find that the drift of
TI away from UT is objectionable, the timekeeping community and ITU
should be able to answer honestly "we told you so".
%%%%%%%%
My scenario requires that the broadcast signals
provide offset ( TI - UT ) as part of their format.
This leaves open the question of at what accuracy level?
I.e., what might happen with leaps of varying sizes?
--> leap microseconds?
number of bits required
to communicate one value almost 48
of ( TI - UT )
amount of history log needed huge. Almost certainly would be
to reconcile UT time tags compressed by piecewise fits to some
with TI time tags functional representation, plus "noise".
effect of leaps on human imperceptibly inconsequential
society and civil events
effect of leaps on timestamping
in information systems probably inconsequential
level of communications probably daily
with IERS required to
keep ( TI - UT ) accurate
to a microsecond
level of effort by IERS considerably more than is currently
required to maintain invested in the system
( TI - UT ) distribution
system
Understanding of the meaning of the UT value being provided by ( TI -
UT ) accurate to microseconds requires deciding what sort of UT1R
smoothing and modelling is being subtracted away -- in effect
requiring all the algorithms in the IERS Conventions and then some
more besides those.
Change of geophysical models completely changes answers that would
have been provided. So logs of UT - TI consist not only of values,
but of algorithms and the rationale behind them.
Burden of this overhead is too great for all but a handful of
specialized systems which already know that they need to make special
arrangements to achieve it.
--> leap milliseconds?
number of bits required
to communicate one value around 32
of ( TI - UT )
amount of history log needed large.
to reconcile UT time tags
with TI time tags
effect of leaps on human inconsequential
society and civil events
effect of leaps on timestamping
in information systems noticeable
level of communications roughly weekly
with IERS required to
keep ( TI - UT ) accurate
to a millisecond
level of effort by IERS approximately equal to what is currently
required to maintain invested in the system
( TI - UT ) distribution
system
Understanding of the meaning of the UT value being provided still
requires deciding what sort of UT1R smoothing and modelling is being
subtracted away.
Change of geophysical models somewhat changes answers that would have
been provided, but logs of ( UT - TI ) still need to footnote the
algorithms then in use and the rationale behind them.
This is no more accuracy for ( UT - TI ) than is currently being
provided in real time over the internet by the IERS, but is more
accuracy than is currently available in the time broadcasts.
It would be quite similar in flavor to the distribution of UT2/UTC
prior to 1972.
--> leap seconds?
number of bits required
to communicate one value around 20
of ( TI - UT )
amount of history log needed same as what we have now.
to reconcile UT time tags It is and will remain manageably large.
with TI time tags
effect of leaps on human just noticeable
society and civil events
effect of leaps on timestamping Some systems are bothered, but most
in information systems don't even set their clocks so well.
level of communications monthly is more than often enough
with IERS required to for the foreseeable future
keep ( TI - UT ) accurate
to a second
level of effort by IERS same as now
required to maintain
( TI - UT ) distribution
system
Basically, providing ( TI - UT ) to one second would be the same as
continuing to provide Rob Seaman's vision of enhanced UTC
indefinitely, though the rules for when ( TI - UT ) would increment
could be changed. But this is less accuracy of UT1 than is currently
provided by time broadcasts.
--> leap minutes?
number of bits required
to communicate one value 16 is plenty
of ( TI - UT )
amount of history log needed Less than what we have now, and
to reconcile UT time tags very manageable.
with TI time tags
effect of leaps on human very noticeable
society and civil events
effect of leaps on timestamping nobody would use UT timestamps,
in information systems they would all be done in TI.
level of communications annually is way more than often enough.
with IERS required to decadally would do for a long time.
keep ( TI - UT ) accurate
to a minute
level of effort by IERS trivially small
required to maintain
( TI - UT ) distribution
system
Leap minutes of ( TI - UT) would give a breather to many of the
systems that supposedly hate leaps while still providing enough
incentive to build in leap handling code and provide a test within the
lifetime (albeit not the working lifetime) of most engineers.
--> leap hours?
number of bits required
to communicate one value 10 is plenty
of ( TI - UT )
amount of history log needed Almost none.
to reconcile UT time tags very manageable.
with TI time tags
effect of leaps on human large, but within limits now accepted
society and civil events by most regional authorities
effect of leaps on timestamping nobody would use UT timestamps,
in information systems they would all be done in TI.
level of communications none, really
with IERS required to We could almost schedule them all
keep ( TI - UT ) accurate now.
to an hour
level of effort by IERS none. Anyone with a stick in the
required to maintain ground on a sunny day can tell
( TI - UT ) distribution when a leap hour should happen.
system
Leap hours mean a Y2K-like readiness review in 500 years.
--> leap days?
No. It's just too much of a break with diurnal rhythms, and it has
day-of-week and related calendrical consequences which are too serious
to contemplate without a flame war.
%%%%%%%%
There are lots more issues that need address, but without knowing
whether the basic assumptions are going to turn out valid it's hard
to be more sure.
--
Steve Allen UCO/Lick Observatory Santa Cruz, CA 95064
sla_at_ucolick.org Voice: +1 831 459 3046 http://www.ucolick.org/~sla
PGP: 1024/E46978C5 F6 78 D1 10 62 94 8F 2E 49 89 0E FE 26 B4 14 93
Received on Sun Jul 06 2003 - 00:14:38 PDT