Re: Leap seconds

[Greg Troxel]

Roger Oberholtzer <Roger.Oberholtzer@ramboll.se> writes:

[formatting repaired]

> When using gpsd/chrony to maintain the local clock, would that time be
> with or without the current 18 leap seconds? In a system where
> everything is pretty much set to defaults.
>
> Our GPS supplier made a comment that surprised me: the NMEA times
> contain the 18 leap seconds. Presumably these times would be used
> unchanged by gpsd/chrony?

Your question uses language that is a bit off, but

  TAI does not have leap seconds.  Only time labs use TAI.

  GPS time does not have leap seconds, but is not the same as TAI.  As I
  understand it, GPS time was coincident with UTC at the start of the
  GPS era, 1980-01-01.  I don't understand why, given that GPS time has
  no leap seconds, it was not set to TAI, which would have saved us from
  one more timescale.  GPS time is used internally within the GPS system
  and is rarely used otherwise.  It can be displayed on some timing
  receivers, but AIUI, only GPS system maintenance people and
  cantankerous nerds would do so.
 
  We live in a GNSS world, not a GPS world, and each constellation has
  its own timescale.  This means that GPS time is even less likely to be
  used in the future.

  GNSS receivers outut NMEA which is in UTC.  UTC has leap seconds
  relative to TAI.  Hence it also has them relative to GPS time.  GPS
  signals transmit a # of leap seconds from GPS to UTC which is how the
  receiver knows how to compute UTC.

  Unix systems keep time in UTC.

  Everything can be messy around a leap second.

So "NMEA times contain the 18 leap seconds" presumably means that the
time reported by NMEA is UTC which is 18s less than GPS time, and thus
"contains leap seconds".   But I would not use the phrase "contains leap
seconds", as it is unclear and unconventional, and I don't think it
leads anybody to understand the situation.

If you are doing normal things, you probably want UTC.  The normal path
of pretty much everything is to use UTC.

> Classification: Confidential

It is nonsensical to publish confidential content to a public list.  In
this case, the only thing disclosed is your lack of clarity about leap
seconds, which puts you in the company of 99.999% of the planet, so
little information has been leaked.

But seriously, please fix your mail system.  It is not reasonable to
send mail with such labels to public lists.