# Leap second functional question - NTP

This is a discussion on Leap second functional question - NTP ; > You are asking for the impossible! Leap seconds keep time in synch with That was the intention. I was pointing out that you can only use "true" time to represent historic civil times, you cannot calculate the "true" time ...

# Thread: Leap second functional question

1. ## Re: Leap second functional question

> You are asking for the impossible! Leap seconds keep time in synch with

That was the intention. I was pointing out that you can only use "true"
time to represent historic civil times, you cannot calculate the "true"
time of a future civil time beyond the first candidate leap second.

> the earth's rotation. The rate of rotation is subject to small
> variations. This is why leap seconds occur at irregular intervals and
> why it is possible to have a negative leap second, although I don't
> recall that we ever had one.
>

Exactly, which is why the question casts doubt on the use of "true"
time, which was my intention.

2. ## Re: Leap second functional question

>>I suppose it is rather unfair of me to point out that I'm a
>>professional astronomer, and can assure you lots of astronomers use
>>UTC.

>
> No very fair. Professional information should trump amatuer every time.
> However, I suspect strongly that when you actually do your calculations you
> use TAI.

What timescale we use depends on the calculation we are doing. There
are times that calculations are done in a TAI like system. There are
times the calculations are done in UTC. UTC is currently kept within a
defined range of UT1, which is very useful.

3. ## Re: Leap second functional question

On 2008-02-20, Unruh wrote:
> No orbital dynamics obeys Newton's ( or Einstein's) law of gravitation only
> in TAI not in UTC ot UT1.

High precision orbital dynamics, such as the JPL's DE405, are done in
Barycentric Dynamic Time, which runs at a different rate than TAI, but
is "TAI like" in that it has no leap seconds.

> Of course I cannot. Leap seconds are to take into account the random
> variations in the earth's rotation to one part in about 10^8.

Leap seconds take into account the systematic slowing of the earth's
rotation.

4. ## Re: Leap second functional question

Unruh wrote:
> "David J Taylor" writes:
>
>> Unruh wrote:

> Note that leap seconds have been happening so rarely that almost noone has
> seen one happen, or watched the system as the clock goes through that.
>

I guess you weren't around in December, 2006.

> Having ntp run on TAI would certainly be simpler, but would of course make
> the time keeping on the system much more complicated.

That question has already been discussed at length in this newsgroup.

Danny

5. ## Re: Leap second functional question

David Woolley wrote:
> 4p4's ntpd/ntp_loopfilter.c:
> /*
> * Set the leap bits in the status word, but
> * only on the last day of June or December.
> */
> tstamp = peer->rec.l_ui - JAN_1970;
> tm = gmtime(&tstamp);
> if (tm != NULL) {
> if ((tm->tm_mon + 1 == 6 &&
> tm->tm_mday == 30) || (tm->tm_mon +
> 1 == 12 && tm->tm_mday == 31)) {
> ntv.status |= STA_INS;
> else if (leap_next &
> LEAP_DELSECOND)
> ntv.status |= STA_DEL;
> }
> }

That comment is wrong. IIRC it can nominally be set at the last day of
any month. The norm is the last day of June or December and in practice
has been only the last day of December. Notice that the code does not
allow a leap second in February which only has 28 or 29 (this year) days
in the month. Either this is a bug or it's already been fixed.

Danny

6. ## Re: Leap second functional question

Danny Mayer wrote:
> }
>
> That comment is wrong. IIRC it can nominally be set at the last day of

The comment matches the code, which should be the most recent released
version. Someone has already said that it is a known bug, and I think
they said it is fixed in the development version.

> any month. The norm is the last day of June or December and in practice
> has been only the last day of December. Notice that the code does not
> allow a leap second in February which only has 28 or 29 (this year) days
> in the month. Either this is a bug or it's already been fixed.

7. ## Re: Leap second functional question

Danny,

That snapshot is older than I thought. That particular botch of code
along with other botches was vulnerable to whimsical errors, like server
leap bits popping up and down or leapsecond file update. The current
code here can leap at the end of any month, which seems to be the
prevailing view of the standards folks. However, I can't find
confirmation at the IERS site.

Dave

Danny Mayer wrote:
> David Woolley wrote:
>
>>4p4's ntpd/ntp_loopfilter.c:
>> /*
>> * Set the leap bits in the status word, but
>> * only on the last day of June or December.
>> */
>> tstamp = peer->rec.l_ui - JAN_1970;
>> tm = gmtime(&tstamp);
>> if (tm != NULL) {
>> if ((tm->tm_mon + 1 == 6 &&
>> tm->tm_mday == 30) || (tm->tm_mon +
>> 1 == 12 && tm->tm_mday == 31)) {
>> ntv.status |= STA_INS;
>> else if (leap_next &
>> LEAP_DELSECOND)
>> ntv.status |= STA_DEL;
>> }
>> }

>
>
> That comment is wrong. IIRC it can nominally be set at the last day of
> any month. The norm is the last day of June or December and in practice
> has been only the last day of December. Notice that the code does not
> allow a leap second in February which only has 28 or 29 (this year) days
> in the month. Either this is a bug or it's already been fixed.
>
> Danny

8. ## Re: Leap second functional question

David Woolley writes:

>> You are asking for the impossible! Leap seconds keep time in synch with

>That was the intention. I was pointing out that you can only use "true"
>time to represent historic civil times, you cannot calculate the "true"
>time of a future civil time beyond the first candidate leap second.

>> the earth's rotation. The rate of rotation is subject to small
>> variations. This is why leap seconds occur at irregular intervals and
>> why it is possible to have a negative leap second, although I don't
>> recall that we ever had one.
>>

>Exactly, which is why the question casts doubt on the use of "true"
>time, which was my intention.

Well, I suppose one could assume no leap seconds into the future. It is as
good as UTC which does exactly that, except it also assumes no leap seconds
into the past, where we know it is wrong. Getting it half right would seem

9. ## Re: Leap second functional question

Greg Hennessy writes:

>On 2008-02-20, Unruh wrote:
>> No orbital dynamics obeys Newton's ( or Einstein's) law of gravitation only
>> in TAI not in UTC ot UT1.

>High precision orbital dynamics, such as the JPL's DE405, are done in
>Barycentric Dynamic Time, which runs at a different rate than TAI, but
>is "TAI like" in that it has no leap seconds.

It does? The second was defined to make the speed of light a consant. Are
you saying they use a system in which the speed of light changes from time
to time? That sounds horrible.

>> Of course I cannot. Leap seconds are to take into account the random
>> variations in the earth's rotation to one part in about 10^8.

>Leap seconds take into account the systematic slowing of the earth's
>rotation.

While it may be slowing on average, it is noise and it could well speed up
as well. For example there have been no leap seconds for the past 4 years
or so, while the "aveage slowing would have one every 1.5 to 2 years right
now. Ie there has been a speedup against the average recently

10. ## Re: Leap second functional question

mayer@ntp.isc.org (Danny Mayer) writes:

>Unruh wrote:
>> "David J Taylor" writes:
>>
>>> Unruh wrote:

>> Note that leap seconds have been happening so rarely that almost noone has
>> seen one happen, or watched the system as the clock goes through that.
>>

>I guess you weren't around in December, 2006.

>> Having ntp run on TAI would certainly be simpler, but would of course make
>> the time keeping on the system much more complicated.

>That question has already been discussed at length in this newsgroup.

And will keep getting discussed since there is no resolution which is
uniformly positive.

11. ## Re: Leap second functional question

Unruh wrote:
>>> Having ntp run on TAI would certainly be simpler, but would of course make
>>> the time keeping on the system much more complicated.

>
>> That question has already been discussed at length in this newsgroup.

> And will keep getting discussed since there is no resolution which is
> uniformly positive.

Actually no. We don't get a vote on this. This is being voted on by the
ITU (or whatever the replacement is for the CCITT) if I recall
correctly. It's a separate question whether or not NTP will continue
with UTC if they do something stupid with the decision.

Danny

12. ## Re: Leap second functional question

On 2008-02-22, Unruh wrote:
>>High precision orbital dynamics, such as the JPL's DE405, are done in
>>Barycentric Dynamic Time, which runs at a different rate than TAI, but
>>is "TAI like" in that it has no leap seconds.

>
> It does? The second was defined to make the speed of light a consant. Are
> you saying they use a system in which the speed of light changes from time
> to time? That sounds horrible.

I am saying that high precision orbital dynamics are done from the
reference frame of the solar system, which we call the barycentric
frame. The barycentric time rate is different from that of TAI by a
factor of about 1e-15 if I remember correctly.

> While it may be slowing on average, it is noise and it could well speed up
> as well.

Noise implies that the changes are both positive and negative, in
which case a leap second won't be needed. If it is systematic,
i.e. the changes are more in one direction than another, a leap second
will be needed.

13. ## Re: Leap second functional question

On Feb 21, 3:15 pm, "David L. Mills" wrote:
> The current
> code here can leap at the end of any month, which seems to be the
> prevailing view of the standards folks. However, I can't find
> confirmation at the IERS site.

Images of the IAU proceedings from the General Assembly in 1973 are at
http://www.ucolick.org/~sla/leapsecs...oc1973p154.gif
http://www.ucolick.org/~sla/leapsecs...oc1973p155.gif
This was the assembly immediately after leap seconds began (and the
first one at which the IAU could make any response, given that the
CCIR neglected to send them a letter in 1970 leaving them no standing
upon which to comment on leap seconds before they began). The changes
recommended by the IAU in 1973 were incorporated into the first
revision of what is now known as ITU-R TF.460, notably, that it was
foreseeable that leap seconds would be needed more than twice a year.

14. ## Re: Leap second functional question

Unruh wrote:

>> High precision orbital dynamics, such as the JPL's DE405, are done in
>> Barycentric Dynamic Time, which runs at a different rate than TAI, but
>> is "TAI like" in that it has no leap seconds.

>
> It does? The second was defined to make the speed of light a consant. Are
> you saying they use a system in which the speed of light changes from time
> to time? That sounds horrible.

I believe he is talking about the general relativity correction. Even
GPS would be wildly wrong if you used the earthbound definition of a
second for the clocks in the satellites.

15. ## Re: Leap second functional question

On Feb 21, 5:30 pm, Unruh wrote:
> >> Having ntp run on TAI would certainly be simpler, but would of course make
> >> the time keeping on the system much more complicated.

> >That question has already been discussed at length in this newsgroup.

>
> And will keep getting discussed since there is no resolution which is
> uniformly positive.

No, but the one I recently espoused in the LEAPSECS group
http://six.pairlist.net/pipermail/le...ry/000190.html
allows for a uniform underlying broadcast time scale without leap
seconds, those being accorded equal status with politically-mandated
time zone shifts, meaning they would go in zoneinfo rather than in the
kernel. It is also in accord with the results of the colloquium on
UTC held in Torino in 2003.

16. ## Re: Leap second functional question

mayer@ntp.isc.org (Danny Mayer) writes:

>Unruh wrote:
>>>> Having ntp run on TAI would certainly be simpler, but would of course make
>>>> the time keeping on the system much more complicated.

>>
>>> That question has already been discussed at length in this newsgroup.

>> And will keep getting discussed since there is no resolution which is
>> uniformly positive.

>Actually no. We don't get a vote on this. This is being voted on by the
>ITU (or whatever the replacement is for the CCITT) if I recall
>correctly. It's a separate question whether or not NTP will continue
>with UTC if they do something stupid with the decision.

I did NOT advocate tht the world run on TAI. I was suggesting that ntp do
so, and that computer systems run on TAI. This makes the translation from
internal time (seconds since epoch) to civil time (UTC) slightly more
complex. Not only does the timezone file need to be consulted, but the
leapsecond file does as well each time you translate from the computer's
internal time to UTC as civil time. UTC is far too uselful for civil time
to be thrown out. Just as the Gregorian calendar reforms were to keep civil
time in sync with the earth's revolution around the sun, so leapseconds
keep the day in sync with the rotation of the earth. Since life is based
around the latter it is good to have time reflect that.
However it is not so good to have the conniptions in the internal time of
the computer try to keep track of that. Seconds since epoch is alrady so
abstract a concept that noone would notice that those seconds were real
seconds, rathr than second where a few a thrown away.

>Danny

17. ## Re: Leap second functional question

Greg Hennessy writes:

>On 2008-02-22, Unruh wrote:
>>>High precision orbital dynamics, such as the JPL's DE405, are done in
>>>Barycentric Dynamic Time, which runs at a different rate than TAI, but
>>>is "TAI like" in that it has no leap seconds.

>>
>> It does? The second was defined to make the speed of light a consant. Are
>> you saying they use a system in which the speed of light changes from time
>> to time? That sounds horrible.

>I am saying that high precision orbital dynamics are done from the
>reference frame of the solar system, which we call the barycentric
>frame. The barycentric time rate is different from that of TAI by a
>factor of about 1e-15 if I remember correctly.

Ah. OK, There are General relativistic corrections, agreed.

>> While it may be slowing on average, it is noise and it could well speed up
>> as well.

>Noise implies that the changes are both positive and negative, in

They are
>which case a leap second won't be needed. If it is systematic,

They would still be needed. Just because your computer's drift rate is both
positive and negative does not mean that compensation is not needed.
They do not necessarily average out on the time scale of years.

There is a net drift to longer days. but superposed on that is a noise
which even over the time scale ofyears makes a difference. An earthquake in
Java rearranges the moment of inertia of the earth and changes the rotation
rate of the earth, and it can be positive or negative.

>i.e. the changes are more in one direction than another, a leap second
>will be needed.

It is needed even if it equal in both directions over a long time.

18. ## Re: Leap second functional question

David Woolley writes:

>Unruh wrote:

>>> High precision orbital dynamics, such as the JPL's DE405, are done in
>>> Barycentric Dynamic Time, which runs at a different rate than TAI, but
>>> is "TAI like" in that it has no leap seconds.

>>
>> It does? The second was defined to make the speed of light a consant. Are
>> you saying they use a system in which the speed of light changes from time
>> to time? That sounds horrible.

>I believe he is talking about the general relativity correction. Even
>GPS would be wildly wrong if you used the earthbound definition of a
>second for the clocks in the satellites.

A common misconception. The GPS people actually dynaically track the time
delivered by the sattelites and adjust their scales accordingly. Even if
they know nothing of GR, they would have discovered that the clocks were
running a bit fast and applied a correction fudgefactor. The problem is
that the clocks on the board could be running at the wrong rate doe to
other environmental problems and they have to be able to correct for those
or your GPS will cease to work.

19. ## Re: Leap second functional question

On 2008-02-22, Unruh wrote:
>>I believe he is talking about the general relativity correction. Even
>>GPS would be wildly wrong if you used the earthbound definition of a
>>second for the clocks in the satellites.

>
> A common misconception. The GPS people actually dynaically track the time
> delivered by the sattelites and adjust their scales accordingly.

It is not a misconception. The tweaks given to the clocks are much
less than the difference between pre launch and on orbit clock rates.

> Even if
> they know nothing of GR, they would have discovered that the clocks were
> running a bit fast and applied a correction fudgefactor.

This sounds like you are saying that even if they didn't know GR, they
would have discovered it. Which may in fact be true, but it doesn't
change the fact that GPS would be wrong if you used the sea level
definiton of a second from the on orbit clocks.

20. ## Re: Leap second functional question

Hello Danny,

On Thursday, February 21, 2008 at 22:01:31 +0000, Danny Mayer wrote:

> The norm is the last day of June or December and in practice has been
> only the last day of December.

There were several leaps in June, the latest in 1997.

Serge.
--
Serge point Bets arobase laposte point net