Re: [Cbor] draft-bormann-cbor-time-tag: POSIX and UTC are not the same

[Carsten Bormann <[e-mail]>]

On 3. Apr 2021, at 04:00, Emile Cormier <[e-mail]> wrote:
> 
> In draft-bormann-cbor-time-tag (tags 1001-1003), UTC and POSIX are confounded together under the same value of 0, under section 3.4. Key -1: Time Scale.
> 
> UTC and POSIX are not the same! Per https://en.wikipedia.org/wiki/Unix_time <https://en.wikipedia.org/wiki/Unix_time>:
> 
> ... leap seconds are ignored, with a leap second having the same Unix time as the second before it, and every day is treated as if it contains exactly 86400 seconds. Due to this treatment Unix time is not a true representation of UTC.

Well, that text is slightly confused, as it is a true representation of UTC, except it can’t represent the 27 leap seconds we have had so far or any of the future ones.

> This is reinforced by the fact that C++ has distinct clocks for system time and UTC:
> 
> https://en.cppreference.com/w/cpp/chrono/system_clock <https://en.cppreference.com/w/cpp/chrono/system_clock>
> https://en.cppreference.com/w/cpp/chrono/utc_clock <https://en.cppreference.com/w/cpp/chrono/utc_clock>
> 
> There should therefore be distinct values for POSIX and UTC.

POSIX and UTC (and thus NTP) are shifted by leap seconds (which is sometimes called “ignoring leap seconds”, a term that should never be used; cppreference says “not counting leap seconds”, which means that you have to be aware of leap seconds to actively not count them).

TAI, GPS are true monotonic time scales.  Leap seconds have no influence on them (what cppreference calls “including leap seconds”).
As C++ utc_clock "includes leap seconds”, it appears to be a TAI referenced time scale (with a constant offset of 10 seconds from TAI).

I’m sorry, I didn’t invent this, I’m just the messenger :-)

Maybe adding an appendix on platform time representations would be worth it…
(We could also add more lines to figure 1, pointing out that C++ utc_clock is TAI minus 10 seconds.)

Grüße, Carsten