Re: [Ntp] Leap second draft

[Warner Losh <[e-mail]>]

On Mon, Apr 1, 2019, 8:27 PM Seaman, Robert Lewis - (rseaman) <
[e-mail]> wrote:

> On 4/1/19, 6:35 PM, Warner Losh wrote:
>
>
>
> The definition of an SI seconds has nothing to do with years, etc.
>
>
>
> If only this were true. The unnecessary confusion of the SI unit of
> interval time (the “second”) with the original definition of the second as
> a fraction of a solar day is what has created the lengthy debate about UTC
> and leap seconds. This then extends to the definition of the solar synodic
> day (
> http://hanksville.org/futureofutc/preprints/files/28_AAS_13-515_Seaman.pdf).
> By choosing the SI unit to masquerade as Universal Time, that is, as a
> fraction of the mean solar day, the precision timing community has
> generated diverse dependencies between what should be independent concepts,
> including between the SI unit and the physical concept of year. That many
> of these dependencies are esoteric does not make them thereby less real.
>
>
>
> By all means the SI second could have been “set by fixing the numerical
> value of the ground state hyperfine splitting frequency of the cesium 133
> atom to be equal to *[say]* exactly *10 000 000 000* when it is expressed
> in the SI unit s–1, which is equal to Hz” (or the corresponding original
> phrasing), but rather the frequency was selected to be *9 192 631 770*
> Hz. This wasn’t a coincidence, but an attempt to obscure the reality that
> atomic time and solar time are two different things. Civil time, and thus
> NTP which conveys it, has dependencies on two different time scales.
>

Historically  the number was chosen to replace the old definition. It is
relevant there. But it's not relevant for the purpose of implementing NTP.

NTP gets it's time from a reference clock, which may care about the details
of the definition, or from a peers, which doesn't. Either way, NTP doesn't
care. It can easily exchange times with a different notion of a second from
SI, so long as all the peers agree to follow the same standard. It could be
GPS which uses SI seconds without leaps. It could have an offset like
exchanging Julian dates. It could even be UT1. This document is not
relevant to any of those. I give them by way of example for why the details
of UTC should be repeated (since they are weird and unique to UTC) and why
the SI second definition need not be (since it will change to be more
precise, but such changes don't matter to NTP since those details don't
affect how you'd write an NTP client or server.)

Warner

> Rob Seaman
>
> University of Arizona
>
>
>
>
>
>
>