[ntp:hackers] ITU and Leap second elimination

Harlan Stenn stenn at ntp.org
Sat Oct 5 10:45:11 UTC 2013


juergen perlinger writes:
> On 10/05/2013 06:08 AM, Mike S wrote:
>> On 10/4/2013 12:51 PM, juergen perlinger wrote:
>>
>>> And are we sure that the leap second smear is something we really need?
>>
>> Absolutely not. That's an abomination. <snip>  At some point, a
>> "smeared" timescale is going to be half a second off from any well
>> defined one. A smear is simply an expedient hack.
>
> That was my first reaction, too, and I have *not* changed my mind on
> that topic.

And I submit that there are a variety of ways to address the issue of
leap seconds:

- use TAI, which doesn't have them
- use UTC, which has them, and then you can:
- - Add a leap second at the rate of 1s/s (Unix) at 23:59:59
- - Add a leap second at the rate of .5s/s (NTP hack for Windows) at 23:59:58
- - Add the leap second using a longer smear

The above are all mechanisms.  Which one anybody chooses is a local
policy choice.

The general timestamp API I've mentioned is designed to handle all of
these cases, identifying them as different "timescales", including a
SMEAR24H, which would smear a leap second over the 86400 seconds before
the leap second.  One could also define other smears if one wanted.

It's arguable that "just add the leap second at the end of the day"
could be SMEAR1S for Unix or SMEAR2S for the Windows hack.

But the thing about the general timestamp library API is that it knows
how to "convert" between these timescales (which include IERS-A, which
undergoes a version update every week) to allow one to properly compare
or convert timestamps.  It should even be able to handle Martian
Standard Time, in which the "day" is longer by about 37 minutes.  NTP
will work just fine with these longer "seconds".

-- 
Harlan Stenn <stenn at ntp.org>
http://networktimefoundation.org - be a member!


More information about the hackers mailing list