Timesys note review

[Rob Seaman]

The vast majority of astronomical data that reference Universal Time
cannot be said to be UT1. If a data product does not trace back to UTC,
then UT is the only appropriate default choice. This is similar to not
preserving more decimal places than supported by input arguments, or not
overreaching on units by reporting microseconds when millisecond or
second accuracy is appropriate.

Indeed, most UT1 values in recent decades would have resulted from
applying the DUT1 correcting to UTC, and would only be precise (and
hopefully accurate) to 0.1 seconds. But more frequently the values
represent the generic notion of Universal Time (UT) and handling of leap
seconds or precision traceability back to a reference clock as would be
needed to claim conformance to the UTC standard are neglected.

Rob

--

On 11/29/18 2:01 AM, Markus Demleitner wrote:
> Hi Steve,
>
> On Wed, Nov 28, 2018 at 08:12:06PM -0800, Steve Allen wrote:
>> On Tue 2018-11-27T13:23:50+0100 Markus Demleitner hath writ:
>>> A pre-built PDF is available on
>>> http://docs.g-vo.org/timesys-draft.pdf, the source is still available
>>> from
>>> https://volute.g-vo.org/svn/trunk/projects/time-domain/timesysnote.
>> Table 1
>>
>> I strongly object to the notion that the term UTC shall be used for
>> time stamps that were originally labelled as GMT.  The term used for
>> that should be just plain UT.
> Ok -- I agree UTC was a bad choice for labelling "historical data
> taken in GMT".
>
> I'd still like to keep the number of (initial) terms as low as
> possible, and I'd expect if we let people choose between UT and UT1,
> that'd not help overall, and I'd still like to escape a separate
> "GMT" time scale if possible.  
>
> So -- given we already have UT1 would you object to saying
> "Historical data given in GMT should be annotated as being in UT1"?
>
>> UTC has always referred to an atomically-regulated time scale.  There
>> [...]
>> The IAU directed that the catalog used for time determination changed
>> from FK3 to FK4 on 1962-01-01, and that caused a shift of 1.5 ms in
>> the values of UT that were being provided by various observatories.
>> The extrapolated estimates of the value of TAI at that date and before
>> have uncertainties and differences of similar size.
> It's clear that automatic processing can only go so far, and we
> should probably set some (time-dependent) goal as to what precision we
> strive for.  I've hoped we could generate some rules of thumb
> analogous to what Arnold has suggested for reference positions, where
> we recommend (overridably) bumping the systematic errors depending on
> time scale and the age of the data.  Your remarks let me doubt it'll
> be so simple... ah well.
>
> Thanks,
>
>           Markus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.ivoa.net/pipermail/voevent/attachments/20181129/d4ae571a/attachment.html>