I think I did address how time measurement is really no different from
weights and temperature..
Here is a message from Mon, 6 Dec 2010 18:55:43 +0100
On Thu, 2 Dec 2010 15:53:40 -0500, Ray Denenberg wrote
> Ed - Your argument for 196u in favor of [1060....1969] certainly
> convinces me that I would rather use an analog than digital
> thermometer if I am seeking accuracy and repeatability. And that I
Digital display technology makes them nearly always more readable than
analog. This is the case for not just measure of temperature, weights
and lengths but also time.
Early mass market battery powered digital clocks from Asia were notoriously
inaccurate. While cheap mains powered clocks used the line frequency (50Hz
60-Hz) these used little more than a simple RC-oscillator--- the first LED
watch, the Pulsar, used a 32.768 Khz crystal or 2^15 but cost $1500 USD
in 1970--- and a counter. LCD made it possible for cheap driver circuits so
there were even some cheap LCD stop-watches sold in the early 1980s that
even offered 1/10th second readability. If anyone here recalls them they
extremely cheap (under $1 USD in Asia) but often lost (I seem to remember
them always running slow) as much as 5 min. day (despite xtal).
Compare now to the readability, repeatability and accuracy of a COSC
chronometer: it can have, at best, 1 second readability but repeatability
better than +-5 seconds/day and accuracy better than +-10 seconds/day
in the extreme case adds up 1 hour/year 365*10 sec = 3650/60 min).
Many wristwatches don't have seconds hands and thus provide readability of
only 1 minute. Some of these have good higher frequency xtal controlled
(CMOS low power) movements and are are thus much more accurate than any
certified chronometer: accuracy up to +-1 min/year (in other words accuracy
and precision within a year are in the same units).
In 1967 the General Conference on Weights and Measures (CGPM) adopted the
atomic definition of the second as the unit of time interval. The second
is defined as the duration of 9192631770 cycles of radiation corresponding
to the transition between two energy levels of the caesium-133 atom.
There used to be something called "Normed Railway Time" (Normalzeit). In
Germany the national railway operated a system of clocks built around Lavet
stepping motors (today the system is still in operation but is now driven
DCF77 radio, e.g. national time reference). These clocks were readable to 1
second. They were accurate (against their own reference) and repeatable to
the speed to electricity to transfer the pulses which put it at under 1
I've mentioned the Hebrew date/time system. Its based upon an astronomical
model: the motion of the Sun. While the Babylonians had a concept of
and sub-seconds (dividing things) the Babylonians measured (and reported
using a double-hour lasting 120 modern minutes, a time-degree lasting four
modern minutes, and a barleycorn lasting 3 1/3 modern seconds. One could
talk about 1/60th of a barleycorn but there was no means to measure it.
This brings us back to the origin of precise and accurate modern second
measurements. This can be credited to Huygens who first suggested the use
of a pendulum, viz. the linkage of time with the force of gravity (constant
at any given spot on the planet).
What we have now is gravity, decay of caesium-133 and the motion of the Sun
across the Meridan.. AND we have measurement of time given my people
some reference clock near an event.. e.g. 1865-05-15T7:22 Washington DC
for the death of Lincoln.
In Judaism there is even geospatial issues for some date/events. Purim, for
example, is celebrated in all places that were surrounded by walls at the
time of Yehoshua as in Shushan (Susa, Iran) on 14 Adar and elsewhere on the
15 Adar. Thus in Jerusalem Purim is 14 but elsewhere the 15th. To
the two one can speak of Purim De'Prazot and Purim De'Mukafot but generally
the indication of day, 14 or 15, is implicitly set by the location of
the speaker or listener in Jerusalem.
I've been talking about measuring the current time. Going the other way..
We often measure the age of things in the lab using Radiocarbon dating.
The interesting thing about carbon dating is that its level of uncertainty
is relatively low and does not increase linearly in time.
The precision of a radiocarbon date is within a few hundred years.
Like the doctors noting what they assume to be the date/time of of
death we also have for Biblical events a chronology specified in the Torah.
The Torah gives us dates with much higher precision than Radiocarbon dating
and the Torah itself can be calculated as being "given" on Mount Sinai 3323
We have another form of dating which I'll call empirical dating. Its
to bibliographic recording. One does not know when, in fact, something was
published but makes a good "educated guess" with a scientific consensus.
dates are not "questionable".. but they may change (in contrast to dates
determined by Biblical chronology) as new states of information and
We will now agree that also for time systems we have differences in
readability, repeatability and accuracy and that higher readability does
not mean higher repeatability nor repeatability mean accuracy-- which
is really a measure against a reference (and includes its own model or
> would rather use a digital thermometer if I don't care so much about
> accuracy and repeatability but put a premium on readability.
> But we're talking about representing a decade, not a temperature,
> and I'm having a hard time seeing the relevance.
> > -----Original Message-----
> > From: Discussion of the Developing Date/Time Standards
> > [mailto:[log in to unmask]] On Behalf Of Edward C. Zimmermann
> > Sent: Wednesday, December 01, 2010 9:30 AM
> > To: [log in to unmask]
> > Subject: Re: [DATETIME] unknown/questionable/uncertain/approximate
> > On Tue, 30 Nov 2010 17:31:45 -0500, Ray Denenberg wrote
> > > Anyway, you are suggesting to merge 'unknown' and 'uncertain', on the
> > > basis that 'unknown' isn't really "unknown" in the sense that 199u is
> > > really "one of [1990, 1991, [UTF-8?][UTF-8?]â€¦., 1999]" so it is
> > > 'uncertain', and in fact both can be represented by a range (as we
> > > define range in the message I
> > posted yesterday).
> > > I think this is a reasonable suggestion. (I am fairly confident that
> > > the
> > "odd" cases, like '1u99' are not real requirements.) I am quite willing
> > to do this (if nobody objects).
> > >
> > I, for one, do object. I don't think we should confuse precisions with
> > ranges set in a higher precision.
> > 196u (aka. the 1950s) is readable (and repeatable) by decade.
> > [1950-1959] might refer also to the 1950s but its clearly readable by
> > year but only repeatable by decade.
> > Using the instrumentation analogy:
> > In instrumentation digital devices tend often to provide much higher
> > levels of readability than repeatability or accuracy. Analog devices,
> > on the other hand, often tend to provide less readability but matching
> > repeatability.
> > Imagine two thermometers. One glass filled with mercury and markings
> > - 00, 10, 20, 30, 50, 50, 60, 70, 80, 90, 100 each 1mm apart from
> > another and the other electronic with a 2-digit display but repeatable
> > to only 10 degree increments.
> > The first thermometer is readable to only 10 degrees.
> > The second thermometer is readable to 1 degree.
> > A bath is measured. The glass thermometer returns the readings:
> > 30, 30, 30, 30, 30, 30, 30, 30, 30, 30 (mercury thermometers offer
> > excellent repeatability) Reading the thermometer I see it looks like a
> > tick above 30 but not really discernible. Clearly not, however, 40.
> > The electronic thermometer returns the readings:
> > 33, 39, 31, 35, 38, 32, 30, 37, 39, 34 (electronic devices tend to
> > drift and have non-linearity resulting in better readability than
> > repeatability) Using ranges it delivered: [30-39]
> > While these may seem effectively "the same" I think they are telling
> > different stories: the decade of the 1950s and the range [1950-1959].
> > Applying now the approximate predicate I can see even different
> > conclusions potentially being drawn..
> > > As to the suggestion that 'approximate' and 'questionable' might be
> > > merged, I am less comfortable. Your interpretation is
> > '"questionable"
> > > is more vague than "approximate"', but I see a more qualitative
> > > difference. The cataloger has some evidence that the event may have
> > > taken place in the year 1150, but no evidence of any other year and
> > if
> > > it wasn't 1150 it could have been - who knows ? Maybe as late as
> > > or even later. One must not infer an approximation, that it it wasn't
> > 1150 then it was sometime close to 1150.
> > > That's a case of '1150?' (questionable). Is that not a meaningful
> > distinction?
> > I wholly agree. We have a number of dates that we assume for things but
> > are widely accepted as questionable. I gave, for example, the biblical
> > Great Flood. There are also dates that are accepted as disputed--- but
> > without any alternative suggestion. Antisa Khvichava, for example, is a
> > Georgian woman who claims as her date of birth 8 July 1880 making her
> > 130 years old. This date is, however, highly disputed. Its suspected
> > that her birth records were either mis-recorded or falsified. Some have
> > even suggested that she might be 20 to 30 years younger than her claim.
> > 1880-07-08 is the only date we have.
> > Its hardly approximate. It might be her date of birth but she might
> > have been born in 1900 or that matter her birthday might not have even
> > been in July or the 8th--- another source of error is the observation
> > that Russia until 1918 used a variant of the Julian calendar.
> > --
> > Edward C. Zimmermann, NONMONOTONIC LAB
> > Basis Systeme netzwerk, Munich Ges. des buergerl. Rechts
> > Office Leo (R&D):
> > Leopoldstrasse 53-55, D-80802 Munich,
> > Federal Republic of Germany
> > http://www.nonmonotonic.net
> > Umsatz-St-ID: DE130492967
On Wed, 28 Oct 2015 14:21:06 -0400, Denenberg, Ray wrote
> The TC154 Committee developing 8601 Part 2 held a conference call
yesterday. Following is a report on several issues of interest to us.
> Use Cases
> Again, my thanks to Hannah. I presented the use case she supplied and
they were all well-received.
> Partial Uncertain/Unspecified
> We had discussion about the suggestion to get rid of the parenthesis
(adopt a different solution) for partial uncertain/approximate, as in:
> 2014-(13)?-12 means only the month (13) is questionable while 2014-13?
-12 means everything to the left of ? , thus both year and month, are
questionable. The potential solutions we had previously discussed were:
> * Use ?? instead if ? to mean only the immediately preceding,
while ? would mean the entire portion to the right (or vice versa).
> * Find another character altogether.
> * Attach ? to each portion that is questionable. Thus for 2014-
13-12? Only the day is questionable, and in order to make the entire date
> None of these solutions is satisfactory.
> However the Chair came up with another idea, which I think is a winner
and the committee accepted: put the ? in front of the questionable portion
to indicate that only that portion is questionable.
> Thus 2014-?13-12 means only the month is questionable. 2014-13?-12
would still mean both the year and month.
> Masked Precision
> In the current EDTF spec, 'u' is a replacement character, or
"placeholder", as in '19uu', used for "unspecified". 'x' is a replacement
character, as in '19xx', used for "masked precision". As expected I have a
difficult time explaining the difference. But I cited the following (from
a 2011 message)
> Imagine, for example, two people that measure the weight of same specimen
using different scales. One person using a good quality analytic lab
balance- -- for example readable to 0.01mg with 0.05mg repeatability--- may
read 1.0146g while the other using a much cruder scale readable to gramms
may read only 1 gram. 1.0146g is not a greater measure than the 1g. Both
have measured the same sample but at different precisions. The weights must
then be "the same". Measuring another specimen the first reads on their
analytic balance 1.0370g. The second person reads 1 gram. At the precision
of the lab balance the second sample is heavier than the first but at the
precision of the crude scale the two weigh the same. Turing now to dates
the date "1964-04" if measured in year precision would yield "1964". In
month precision "1964-04" is before "1964-06" but in year precision they
are "the same" just as two babies born on the same day but at different
times share the same birthday.
> The scale example is a good explanation of precision. But the leap to
dates is hard to grasp. When you say 'Turing now to dates the date "1964-
04" if measured in year precision...' , there is an implied "scale" analogy
that measures time rather than weight. We need to have a better
understanding of how time is being measured.
> Long Year - Exponential Form
> Recall there was discussion among us to eliminate the exponential form,
though to retain the precision component. Thus instead of
> y17101e4p3 (Some year between 171000000 and 171999999, estimated to be
171010000 - 'p3' indicates a precision of 3 significant digits.)
> It would be
> However, the committee prefers to retain the exponential form.
> Furthermore, 'P' cannot be used, as it is already used within 8601 (for
"period"). Instead, 'A' is suggested (for "accuracy").
> It really would be useful to be able to specify a century (e.g. 20th
century) unambiguously, even if only within the context of the
specification. That is to say, the spec should define precisely what
"century" means, while making it clear that the definition is not intended
to define "century" for all mankind, but only for use within the spec. It
is clear that the string '19' is intended to mean 1900-1999, but 8601 never
says so. It supplies '19' as an example of "a century" but doesn't say
explicitly what period it represents.
> I proposed the following definition, which the committee accepted:
> "A century is represented by two digits, 'nn' (meaning the [nn+1] th
century) which for purposes of this standard signifies the one-hundred year
period consisting of all the years nnxx where xx is any two-digit number.
Thus for example '19' is the 20th century, and represents the 100 year
> Season - Qualified
> The season qualifier http://www.loc.gov/standards/datetime/pre-
submission.html#seasonqualified will be dropped. Instead, the numbering -
21,22,23,24 - will be extended as necessary for seasonal expressions. E.g.
perhaps 25 will mean "Spring - Northern Hemisphere". There will be an
initial set in the published standard and a mechanism to register
> L1 and L2 Features
> With the possible exception of masked precision (which is still up in the
air) all level 1 and level 2 features are accepted by the committee, though
with various changes (that have been noted). That means they will be part
of the proposed standard; it does not mean that they will be accepted by
ISO in general.
Edward C. Zimmermann, NONMONOTONIC LAB