Re: [Ntp] Leap second draft

On Mon, Apr 1, 2019 at 6:35 PM Warner Losh <imp@bsdimp.com> wrote:
>
>
>
> On Mon, Apr 1, 2019 at 7:14 PM Watson Ladd <watsonbladd@gmail.com> wrote:
>>
>> On Sat, Mar 30, 2019 at 8:02 PM Warner Losh <imp@bsdimp.com> wrote:
>> >
>> >
>> >
>> > On Sat, Mar 30, 2019, 6:36 PM Daniel Franke <dfoxfranke@gmail.com> wrote:
>> >>
>> >> On Sat, Mar 30, 2019 at 7:53 PM Kurt Roeckx <kurt@roeckx.be> wrote:
>> >> > Anyway, I think we should avoid defining what things like NTP
>> >> > time, UTC and TAI is. They should all have external definitions
>> >> > we can just point to.
>> >>
>> >> Where the definitions are normative, sure. But as Steve's infodumps in
>> >> this thread rather vividly illustrate, readers need a comprehensible
>> >> summary; we can't just incorporate the whole body of relevant
>> >> standards by reference and leave it at that. The statements in the
>> >> background section aren't normative, and should be accurate but don't
>> >> have to be comprehensive. And I won't be worried if, for example, the
>> >> quoted definition of an SI second gets superseded a month after the
>> >> RFC is published, because the quote isn't there to be authoritative;
>> >> it's there to illustrate what such a definition can look like.
>> >
>> >
>> > The current definition doesn't help. A reference to the BIPM with the latest definition doc at the time of publication would suffice. The hard part is getting the right pointers.
>> >
>> > It doesn't matter what a SI second is. NTP counts them and exchanges that count. They could be anything. People have done GPS times (also SI seconds) and UT1 times (solar variable length seconds). They've even done Julian Time NTP.. Now obviously these different time scales won't interoperate, but they are fine so long as everybody exchanging time agrees on the rules for the time scale. That brings us to UTC. It's the odd man out. It ticks at a fixed rate, and occasionally has seconds inserted into it. UTC is related to TAI by a stepwise function that always increases or decreases by 1. This increase or decrease historically has always been December or June, but technically could be at the end of any month. UTC is synchronized to UT1 every so often based on a set of rules. The current rule is that | DUT1 | must remain < .9s so you move it when you think it might exceed .9 before the next chance to move it. Other rules are possible, but just like leaps at the end of any other month than December or June not likely.
>>
>> I've tried citing the BIPM definitions, but they are specified in
>> years, months, days, hours, minutes, seconds, and fraction of second,
>> while NTP uses seconds and fraction of seconds. Then there is the
>> whole document access issue.
>
>
> The definition of an SI seconds has nothing to do with years, etc.

Correct, but UTC timestamps are specified with a representation based
on hours, minutes, and seconds.

>
> Warner
>
>>
>> >
>> > So, it makes sense to just hit these points, and give pointers to the underlying details, which don't matter.
>> >
>> > Warner
>> >
>> >> _______________________________________________
>> >> ntp mailing list
>> >> ntp@ietf.org
>> >> https://www.ietf.org/mailman/listinfo/ntp
>> >
>> > _______________________________________________
>> > ntp mailing list
>> > ntp@ietf.org
>> > https://www.ietf.org/mailman/listinfo/ntp
>>
>>
>>
>> --
>> "Man is born free, but everywhere he is in chains".
>> --Rousseau.

-- 
"Man is born free, but everywhere he is in chains".
--Rousseau.