On Mon 2021-04-26T17:17:13+0100 Tony Finch via tz hath writ:

My aim back then was to sketch what might be a reasonable way to bridge between pre-atomic time and UTC, for code that can only support an integer offset between UTC and TAI. If your code supports rubber seconds then it should probably use the USNO table that records the rate and phase offsets between 1962 and 1972. (I don't know another source of that table, nor where you can get a copy while maia.usno.navy.mil is down.)

Keep in mind that the USNO numbers only apply to time signals broadcast in the US, and before 1968 they ignore the many microseconds of offset between NBS and USNO. Germany did not start to broadcast the old UTC until the late 1960s, until 1974-01-01 broadcast both old UTC and SAT, and stopped providing old UTC as of 1970-04-01 because it had been deemed illegal. Soviet and Chinese time signals never used the old UTC. Sussing out the particulars for individual cases of national time signals needs finding remnants of the high-acid paper that each national time service used for bulletins issued to its users, and/or diving through the BIH Bulletin Horaire and Circular D to pull out the received timing of each official broadcast. It is much simpler to ignore the 50 to 200 ms jumps that were put into old UTC and say that before 1972 all available sources of civil time provided UT (nevermind the particular flavor, nobody set their clock that accurately) and after 1972 UTC with leap seconds. For civil time I do not see any advantage in a model with greater complexity than just plain UT as the basis time scale before 1972-01-01. This is adequate for any IANA tz library. For precise time no general model applies. Each precise timestamp must be meticulously referenced to the particulars of the clocks that were used to obtain it. No available clocks were using TAI before 1972-01-01, and this does not belong in any IANA tz library. Constructing a model that supposes it is possible to do better is akin to writing The Silmarillion. Look at the BIH plots of the accuracy of clocks https://www.ucolick.org/~sla/leapsecs/BHsHn05p142.html millisecond agreement between different national authorities was not achieved until 1960 during the UK/US experiment that became old UTC.

* TAI is a retrospective timescale; it isn't defined in the future.

TAI also cannot be defined prior to mid 1955, and it was not generally availble for use by any system until 1972-01-01 when broadcast time signals began to be based on TAI. The BIPM will never tolerate TAI becoming part of an international agreement. The best that can be hoped for in a new international agreement on time is one (or more) new TAI-like uniform scale which will be offset by some integer number of seconds from TAI. The reason we use UTC is because it is what is in the current international agreement. Everybody involved in the technical aspects eschewed the use of UTC for their systems from before 1972-01-01. Look at the ATSC agreement for a technically-competent time scale. They chose GPS time because it is practically available. ATSC was not bound by international bureaucratic requirements, so they could make that choice. Because it is ultimately controlled by USNO GPS system time will never become part of an international agreement. On Mon 2021-04-26T12:42:19-0400 John Sauter via tz hath writ:

In looking at historical estimates of the rotation of the Earth based on astronomical observations of occulations and eclipses, I found that the values of UT1 are reliabe only since 1825. (Nevertheless, my table of proleptic leap seconds goes back to the year -2000.)

Morrison, L. V.; Stephenson, F. R.; Hohenkerk, C. Y.; Zawilski, M. DOI: 10.1098/rspa.2020.0776 Bibliographic Code: 2021RSPSA.47700776M Proceedings of the Royal Society A, Volume 477, Issue 2246, id.20200776 all tabulated at http://astro.ukho.gov.uk/nao/lvm/ That is the place to look, and it will continue to be so for the lifetime of the authors. But what is it mean to ask reliable? That is very much the question given that prior to "reliable" everybody's clock was set to local apparent time. What exactly is the question that is trying to be answered by an algorithm like this? Ultimately it is necessary to model two kinds of time: 1) earth rotation time (which before the 20th century differed by a second or more depending on who was setting the clock) because that is a subdivision of internationally agreed calendar days 2) uniform time (which becomes possible to interpret as atomic time beginning in 1955). -- Steve Allen <sla@ucolick.org> WGS-84 (GPS) UCO/Lick Observatory--ISB 260 Natural Sciences II, Room 165 Lat +36.99855 1156 High Street Voice: +1 831 459 3046 Lng -122.06015 Santa Cruz, CA 95064 https://www.ucolick.org/~sla/ Hgt +250 m