[LEAPSECS] Computer Network Time Synchronization, 2nd Ed.

[Rob Seaman]

On Dec 7, 2011, at 12:48 AM, Poul-Henning Kamp wrote:


>> I wonder where he got the idea that death and birth certificates are UT1 too.

> 

> Or for that matter, what difference it could possibly make ?


Nice way to ignore the fact that you chose to excise the one key term "UTC" from Mills' quote.


> Birthdays do not even get adjusted with/for timezones: You are born and die on civil time,


Oh please.  UT1 (~GMT ~UT ~Universal Time ~mean solar time ~time-of-day ~synodic time) is the basis of the entire time zone system.  (That's why there are so many synonyms for the same concept.)  Obviously some of us disagree with your assertion that none of that matters or that the ITU is free to spin the dial like a pointer in a board game.  If you want to talk constructively about tweaking the approximation, go right ahead.  Attempting to throw the entire concept out is nonsense.


>> [...]be registered immediately after birth..."  Just the fact that

>> the Convention says "a child means every human being below the age

>> of eighteen years" creates engineering requirements on dates and

>> thus on time.

> 

> Check the pass-port convention, can't remember the name of it,

> possibly the Warshaw-Convetion (UN has a open database with all the

> international conventions)  I belive it only strongly urges birthday

> to be recorded in passports by the day, but only requires them to be

> exact to a whole month.  Not sure if the E-passport is a convention

> or just treaties.


Interesting pointer, thanks.*  As usual, however, you confuse precision with accuracy.  The fundamental error with the ITU proposal (whatever it actually says these days) is not an error in offset, it's the perpetual rate error.  It is precisely that dreaded quadratic drift that is being introduced into civil timekeeping, not avoided.


>> "UT1", is - of course - just an indication that *actual* time-of-day

> 

> Why don't you ask Dave Mills, rather than fit your agenda to his opinions ?


Ah - I see!  Just a slip of the tongue, was it?  Why Rob!  Thanks for the erratum; what I meant to say wasn't:

	"In the end, it seems prudent that the computer clock runs in UTC with leap insertions as described."

But rather:

	"I encourage the ITU to act immediately to redefine UTC so that our long international nightmare can end!"

One might tend to trust the carefully chosen words of a newly revised and expanded book describing his life's work.  Strangely, you don't.

As far as making words fit, I note that you yourself chose to focus on just one sentence in the middle of the five sentence paragraph I quoted.**  And still had to cut out the smoking gun "UTC" to make it fit your own agenda.

But why so touchy if delegates at the assembly in Geneva are just going to wave their paddles all day long for any measure that comes before them?

Rob
--

* Of course, months are another synodic concept.  One that was redefined prior to the computer age.  One wonders if otherwise we would be having the same conversation now regarding months in addition to days.  Whatever the relationship between human society and the synodic rhythms of the quirkily imperfect clockwork solar system we inhabit, there most definitely is a line that we can't cross in pretending that the Sun doesn't exist.  For instance, calendar months would make an unacceptably poor clock on the Moon, where the synodic month is the synodic day (since Luna is tidally locked).

** Full disclosure, the rest of the paragraph I omitted from the quote was:

	"By agreement between USNO, NIST, and the NTP developer community [8], the NTP Version 4 (NTPv4) Autokey protocol described in Chapter 9 has been modified to auomatically download the leap second table from NIST.  Assuming the operating system kernel has the required capability, the leap insertion is implemented automatically at the required epoch and the current TAI offset made available via the kernel application program interface."

Reference [8] is:  Levine and Mills, "Using the Network Time Protocol to transmit International Atomic Time (TAI)", Proceedings Precision Time and Time Interval (PTTI) Applications and Planning Meeting, Reston, VA, November 2000.

Even the name "Precision Time and Time Interval" makes the distinction between the two different types of time scale, between clocks and chronometers.  In any event the problem appears to be a solved one.  Perhaps part of the artificial urgency of this issue is to make it a moot point before NTPv4 is universally deployed?