On 2014-01-17 04:06 AM, Zefram wrote:
- "correction" is more widely understandable than "compensated" or
some more technical term.
"Correction" suggests a term that must always be added into certain
computations. This is quite different from a leap, which is an one-time
irregularity in an otherwise regular sequence. To be fair, what your
table labels "Earth Correction" does look like a correcting term.
Well, I suggested what I felt was a reasonable "plain speak" term,
hoping to make it more understandable to non-technical people. I guess
technically "compensation" is better, but no name is going to be
prefect. I'm open to suggestions.
Your table is incorrect in showing `UTC' times prior to 1961 and a `TAI'
time prior to 1955. As discussed above, TAI is not defined for dates
prior to the use of atomic clocks. UTC is likewise not defined prior
to the international agreement on how to steer time signals relative
to TAI. It's legitimate to define some time scale that is continuous
with TAI and is defined for 1900, and you can then describe a 1900 time
on that time scale, but if you refer to actual TAI then you are subject
to its limitations.
Sure, we need clear definitions, thats the idea of finding agreement.
The best I'd thought of so far was "Proleptic TAI" and "Proleptic UTC",
but I agree those concepts along that portion of the timescale may want
their own names.
The point, as explained in earlier, email is to intentionally sweep the
history under the rug prior to 1972-01-01T00:00:00Z.
>From 1961 to 1971 the correspondences you show between UTC and
TAI are not correct. You have written the table as if UTC were
defined as exactly TAI - 10 s prior to 1972, which is not the
case. The difference was 10 s only for the first half of 1972.
See <http://toshi.nofs.navy.mil/ser7/tai-utc.dat> for the expressions
defining rubber-seconds UTC.
Yes, I understand that very well. Again, sweep the history under the rug
prior to 1972-01-01T00:00:00Z.
You show your "Earth Correction" being a constant 10 s prior to
1972, and following the TAI-UTC difference thereafter. This makes a
poor correction.
Yes, thats true. The point there is to establish consistent counting
methods for computer time-keeping, keyed off NTP and POSIX.
NTP clearly has 10 Leap Seconds in effect at its "1900" "prime epoch" on
the "Proleptic UTC" timescale.
POSIX also has 10 Leap Seconds in effect at its "1970" "the Epoch" when
treated on the "Proleptic UTC" timescale, as NTP indicates it should be.
CCT could explicitly state this.
Neither NTP nor POSIX can be accurate with respect to UTC before 1972.
If the intent is to define a proleptic version of
modern UTC, you need to decide on dates for proleptic leap seconds.
For continuity your Earth Correction needs to at least approximately track
the TAI(TT)-UT1 difference. (By "TAI(TT)" I mean TT - 32.184 s; that is,
TT adjusted to the TAI epoch, making a fake TAI that can be extended back
proleptically prior to 1955. I presume this to be a good approximation
for anything that would match your "essentially proleptic TAI" intent.)
The main objective of CCT as far as I had gone was define a timescale
that encompassed deterministic computer-based time-keeping. If a
"proleptic UTC" scale is used as I've suggested, it will be increasingly
inaccurate to the actual date-time before 1972 depending on exactly how
Leap Seonds (Earth Correction) is applied prior to 1972.
There may be need for, or interest in, defining "proleptic Leap Seconds"
back to, say, 1958 or 1955, to cover the historical record. I think
tracking the non-integral steps and frequency changes of SI during this
period may be of little practical value, but there may be good reason
for it for some purpose I'm unaware of.
I think it would useful, or at least interesting, to define "proleptic
Leap Seconds" into the past, perhaps into the deep past. Steve Allen
shows a graph of "Ancient values for Length of Day" at
http://www.ucolick.org/~sla/leapsecs/dutc.html. By assigning Leap Second
values as best can be determined you'd have a pretty good timescale for
historians.
-Brooks
_______________________________________________
LEAPSECS mailing list
LEAPSECS@leapsecond.com
http://six.pairlist.net/mailman/listinfo/leapsecs
