UTC
Poul-Henning Kamp says:
Other people think that by sheer majority of users, the UTC
timescale no longer belongs to the astronomers, and that it can and
should be redefined to serve its major target audience better.
You really need to get out more...
I did get out more. I've been focusing on two other issues entirely
since scribbling that message. In the week or so since the message I
was replying to was written, there have been dozens of other things
going on. I will continue to return to UTC since it is an ongoing
problem. Glancing over my message, I regret deciding to throw
together a response - but it is hard to find time to give every issue
the time it deserves to polish one's responses.
That said, are you of the opinion that your limited perception of my
project list (or social calendar, for that matter) has anything at
all to do with resolving technical issues? As you say, some fraction
of the people interested in the topics of the definition and
transport of civil time believe that UTC is a throwaway time scale
that should be sacrificed for the greater good. Believing this does
not make it so.
Let me reiterate my core assertions:
-
Civil time is a subdivision of the calendar and is thus a
representation of time-of-day. -
That said, one could imagine basing civil time on some time scale
other than UTC. This is a policy question that should be addressed
forthrightly, not via a tricky initiative involving a mid-level
bureaucratic committee. We aren't just talking about leap seconds,
we are talking about the definition of the concept of the "day".
This is a potentially explosive political, legal and religious
issue. Its current obscurity should not be taken for granted. -
Whatever time scale civil time is based upon, the integrity of the
underlying time scale should be respected. UTC has an existence
separate from the ITU. It should be maintained whether or not civil
time continues to correspond to UTC. -
The real issue is the improvement of systems, data structures, and
APIs used to transport time signals corresponding to multiple time
scales. The current proposal is as much an attempt to sweep this
important work under the rug, as it is an attempt to avoid subjecting
non-conforming systems to the horrors of leap seconds. Shame on
anyone who is looking for an easy way out.
Time is complex. Embrace the complexity.
Rob Seaman
NOAO
In message 75DA592E-7563-4B0C-A418-4148590A7EAF@noao.edu, Rob Seaman writes:
Poul-Henning Kamp says:
That said, are you of the opinion that your limited perception of my
project list (or social calendar, for that matter) has anything at
all to do with resolving technical issues?
No, I'm saying that you thinking it is a technical issue is pretty
much the central problem in our discussion.
This is a political issue because the economics of following
the "obvious technical solution" are prohibitive.
- Civil time is a subdivision of the calendar and is thus a
representation of time-of-day.
The almost universal adoption of timezones and daylight savings
time disproves this definition: Civil time varies up to 30
degrees or two hours from "time-of-day" in many areas of the
world.
- That said, one could imagine basing civil time on some time scale
other than UTC.
Yes, one could conceiveably imagine that.
But only if one ignores the the astronomical cost of converting all
the documents, manuals, directives, schedules, agreements, contracts,
trading rules, flight schedules, (just to mention some) of the
places that now have text that refer to UTC.
For people who are aware of this cost, it is immediately obvious
that saying "Make your own damn timescale" to the astronomers is
so many orders of magnitude cheaper, that such a proposal will never
be taken serious in intelligent debate.
This is a policy question that should be addressed
forthrightly, not via a tricky initiative involving a mid-level
bureaucratic committee.
I can guarantee you that it has been addressed definitively
in my statement above. You will not ever get another decision
anywhere.
We aren't just talking about leap seconds,
we are talking about the definition of the concept of the "day".
No, we are not.
A day is defined as 24 hours and the fact that some of them are
24h1s is in violation of countless laws and regulations throughout
the world.
This is a potentially explosive political, legal and religious
issue. Its current obscurity should not be taken for granted.
Right, and we don't: we're trying to get rid of the 24h1s obscurity.
- Whatever time scale civil time is based upon, the integrity of the
underlying time scale should be respected. UTC has an existence
separate from the ITU. It should be maintained whether or not civil
time continues to correspond to UTC.
If anybody put their eggs in the UTC basket and didn't realize
that ITU carried it, they have only themselves to blame.
The current proposal is as much an attempt to sweep this
important work under the rug, as it is an attempt to avoid subjecting
non-conforming systems to the horrors of leap seconds. Shame on
anyone who is looking for an easy way out.
Wrong.
When cost estimates start running into the billions of dollars,
looking for an easy way out is not optional any more, then it
becomes a duty.
And saving even a few tens of millions of dollars can never be
a shamefull thing for anybody.
Time is complex. Embrace the complexity.
No, geology and orbital mechanics is complex.
Time is just as simple as any other fundamental measurement unit
like the meter, the kilogram or the volt.
Once you stop confusing geophysics with time, then we will be
able to discuss this rationally. As long as you confuse the
two, we get nowhere.
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
I think I will trouble the list with just one more message on this
topic, as I'm in danger of repeating myself - certainly the folks I'm
replying to have started to do so. It is also remarkable how many
ways there are to make my arguments for me. Too expensive to revise
documentation? So you're saying that a major change to civil time
will be instituted such that UTC no longer approximates GMT and
nobody should be told about it? Worse yet, that the documentation
that does exist will be allowed to lie?
Poul-Henning Kamp says:
The almost universal adoption of timezones and daylight savings
time disproves this definition: Civil time varies up to 30 degrees
or two hours from "time-of-day" in many areas of the world.
All true. All irrelevant.
Pop quiz! What is the length of the day? No tricks - no gimmicks.
Launch a tee-shirt to that guy in the last row. Right you are! A
day on Earth is 23h 56m 4s. Normal SI seconds. And a simple 360
degree rotational period with respect to a reference grid of fixed
extragalactic sources.
PHK confuses periodic effects (such as the analemma traced out by the
equation-of-time) with secular effects - a monotonically diverging
DUT1 in particular. There are many co-existing solar time scales:
local apparent time, local mean time, standard time, daylight saving
time, and yes - Coordinated Universal Time to tie all the localities
together. This multiplicity of time scales does not weaken the
argument for having UTC continue to track time-of-day - it
strengthens it. Universal Time is the modern formulation of the
concept of Greenwich Mean Time that has served us so well. UTC is
that concept joined to International Atomic Time via leap seconds.
It is only the fact that UTC continues to approximate GMT that brings
sanity to the long tally of solar time scales.
It is very telling that arguments such as PHK's always seem to be
based on solar time. After all, the spinning Earth forms a steady
clock relative to the fixed stars, not the Sun. The length of the
day "really, truly" is 23h 56m 4s. Why is this never pointed out?
Because it simply isn't so for most purposes. A day in the life is a
solar day. It is the solar day our clocks should characterize. That
the Earth's orbit is elliptical, its axis tipped, and that it is
circled by the Moon simply underscore the need for civil time to
follow mean solar time and thus bring clarity to our timekeeping. It
is precisely the overwhelming dominance of the Sun in our everyday
lives that illuminates the clear choice of basing civil time on solar
time. It is just as wrong for the precision timing community to
embrace the split second (but growing) errors of raw TAI as it would
be to permit the nearly four minute daily error that sidereal clocks
would impose on our lives.
Rob Seaman
National Optical Astronomy Observatory
In message 4A4B4876-7D7F-4D81-AE31-26A757E7AA81@noao.edu, Rob Seaman writes:
Pop quiz! What is the length of the day? No tricks - no gimmicks.
Launch a tee-shirt to that guy in the last row. Right you are! A
day on Earth is 23h 56m 4s.
Now, for two t-shirts:
Which fraction of the earths population would disagree with the answer
that astronomer gave ?
Right there in front: 99.999% sounds about right.
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
Poul-Henning Kamp wrote:
In message 4A4B4876-7D7F-4D81-AE31-26A757E7AA81@noao.edu, Rob Seaman writes:
Pop quiz! What is the length of the day? No tricks - no gimmicks.
Launch a tee-shirt to that guy in the last row. Right you are! A day on Earth is 23h 56m 4s.
Now, for two t-shirts:
Which fraction of the earths population would disagree with the answer that astronomer gave ?
Right there in front: 99.999% sounds about right.
Now, for 3 t-shirts Poul, reread Rob's last paragraph, but this time with your blinders removed.
[-----Rob Seaman's last paragraph-----]
It is very telling that arguments such as PHK's always seem to be based on solar time. After all, the spinning Earth
forms a steady clock relative to the fixed stars, not the Sun. The length of the day "really, truly" is 23h 56m 4s.
Why is this never pointed out? Because it simply isn't so for most purposes. A day in the life is a solar day. It
is the solar day our clocks should characterize. That the Earth's orbit is elliptical, its axis tipped, and that it is
circled by the Moon simply underscore the need for civil time to follow mean solar time and thus bring clarity to our
timekeeping. It is precisely the overwhelming dominance of the Sun in our everyday lives that illuminates the clear
choice of basing civil time on solar time. It is just as wrong for the precision timing community to embrace the
split second (but growing) errors of raw TAI as it would be to permit the nearly four minute daily error that sidereal
clocks would impose on our lives.
[-----Rob Seaman's last paragraph-----]
You can agree to disagree, but Rob was using the 23h 56m 4s sidereal day as a parable, not as an actual
suggestion that we all abandon solar/seasonal time.
Neither of you is as dumb as you think the other to be.
-Chuck
I feel obligated to reply lest someone think Poul-Henning Kamp scored
a point off me:
Pop quiz! What is the length of the day? No tricks - no gimmicks.
Launch a tee-shirt to that guy in the last row. Right you are! A
day on Earth is 23h 56m 4s.
Now, for two t-shirts:
Which fraction of the earths population would disagree with the
answer that astronomer gave ?
Right there in front: 99.999% sounds about right.
This was, of course, exactly my point. We wouldn't even be having
this discussion if the length of an SI second hadn't been chosen to
approximate a "solar" second. Four minutes per day would certainly
be too much to tolerate - but so would 1s per day. Some may think
that solar time is a fantasy - but the length of a solar day forms a
very precise limit on civil timekeeping.
One could - and often does at a telescope - use a sidereal time scale
in which 24 hours of 60 minutes of 60 seconds passes in the time that
it takes for the Earth to rotate once with respect to the stars. One
sidereal day of 86400 sidereal seconds. Few other than astronomers
use such days or such seconds. (Although sidereal signals are easily
detected at radio frequencies.)
Similarly, only a handful of engineers, programmers and scientists
use what might be called "atomic" days.
Everybody (100.000%), however, uses solar days. And the cleverness
of UTC is that they are able to use atomic (SI) seconds at the same
time. The best of both worlds often comes with a price. That price
is the leap second.
The ITU proposal isn't a challenge to leap seconds issued every few
hundred days. The proposal is a challenge to the integrity of each
day that dawns. And if such a major change is to be made to a
fundamental international standard, the process used to decide this
new policy should represent the best of international cooperation -
not the sneakiest.
Anybody want to try for three tee-shirts?
What fraction of the Earth's inhabitants believe that a day is
defined to be a solar day?
Rob Seaman
National Optical Astronomy Observatory
In message 42E6C1F7.3040002@erols.com, Chuck Harris writes:
Poul-Henning Kamp wrote:
In message 4A4B4876-7D7F-4D81-AE31-26A757E7AA81@noao.edu, Rob Seaman writes:
Now, for 3 t-shirts Poul, reread Rob's last paragraph, but this
time with your blinders removed.
I'm still trying to get the same basic point across, and answering
his last paragraph would just have repeated the message yet again:
To 99.999% of the Earths population, the Earth is geophysics, not
timekeeping.
For almost the entire population, time is defined as SI seconds,
60 of which is a minute, 24 of which is an hour, 365 of which is a
year.
There is a well defined formula that streches indefinitely into the
future, which inserting an extra day almost every fourth year, for
reasons most people do not claim to fully understand.
I fully agree with Rob that there are prefectly valid physical and
astronomical reasons for the leapsecond, but I keep trying to tell
him that like so many other metrological details, the everchanging
mass of the kilogram prototype comes to mind, it will have to be
a dirty little secret in the scientfic world: We cannot economically
justify imposing leap seconds on the rest of the world with six
months notice.
Leap days are economically feasible because most people have a big
disconnect between the concepts of timekeeping and calendar but
mostly because the rule is predictive and can be embedded safely
in computers which are going to run autonomously for years.
Leap seconds are too small for programmers to have heard about or
care about, they are unpredictable for more than six months at a
time and untestable in practice.
Robs position, as I understand it is "fine, fine, fine! just don't
call it UTC!" and that is not economically feasible either IMO.
Leapseconds is not a scientific or astronomical issue, they are
an economical issue.
Poul-Henning
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
Poul-Henning Kamp wrote:
In message 42E6C1F7.3040002@erols.com, Chuck Harris writes:
Poul-Henning Kamp wrote:
In message 4A4B4876-7D7F-4D81-AE31-26A757E7AA81@noao.edu, Rob Seaman writes:
Now, for 3 t-shirts Poul, reread Rob's last paragraph, but this
time with your blinders removed.
I'm still trying to get the same basic point across, and answering
his last paragraph would just have repeated the message yet again:
To 99.999% of the Earths population, the Earth is geophysics, not
timekeeping.
For almost the entire population, time is defined as SI seconds,
60 of which is a minute, 24 of which is an hour, 365 of which is a
year.
That is correct, but you must also remember that 99 and 44/100ths of
the population would do just fine with a timepiece that had no second
hand. Seconds are just a frill to civil timekeeping.
If you have been watching the population in general, you may have noticed
that there is a decline in the number of people that have any interest in how
things work. When I was a kid, there wasn't a toy, or appliance that was
safe from my curiosity. I suspect that you were the same. Now, with the
exception of sadistic impulses that remove the heads from dolls, kids don't
take their toys apart. They just want them to work, and when they stop, they
throw them away.
The point being, the needs of the population-at-large is not a valid reason for
making a decision as to whether we need leap seconds, or don't. The population-
at-large just doesn't care, at all.
That puts the decision down to the very few that do care. Those that care have
already boiled the time situation down to two systems which are each synchronized
to the atomic second, TAI, and UTC. UTC was designed to be a replacement
for GMT. For that to be true, it needs to remain synchronized to solar time
at the zero meridian. That, regardless of how we feel about the implementation,
requires a leap second from time-to-time.
If UTC ceases to remain synchronized to solar time at the zero meridian, then
how does it differ from TAI? It becomes redundant, and may as well cease to
exist.
There is a well defined formula that streches indefinitely into the
future, which inserting an extra day almost every fourth year, for
reasons most people do not claim to fully understand.
They don't need to understand it, they just need someone to tell them
that that is the way it is, and they are happy. Don't forget, they don't
care about the details of how things work.
I fully agree with Rob that there are prefectly valid physical and
astronomical reasons for the leapsecond, but I keep trying to tell
him that like so many other metrological details, the everchanging
mass of the kilogram prototype comes to mind, it will have to be
a dirty little secret in the scientfic world: We cannot economically
justify imposing leap seconds on the rest of the world with six
months notice.
Certainly we can. As I said earlier, 99 and 44/100ths percent of the
population has no need for seconds whatsoever. The remaining 56/100ths
can slip a leap second in when one is required. And they can do so
when they get around to it, or not at all, and there will be little or no
consequence. For those that cannot handle the ambiguity presented
by the leap second, there is TAI.
Leap days are economically feasible because most people have a big
disconnect between the concepts of timekeeping and calendar but
mostly because the rule is predictive and can be embedded safely
in computers which are going to run autonomously for years.
Leap seconds are too small for programmers to have heard about or
care about, they are unpredictable for more than six months at a
time and untestable in practice.
Robs position, as I understand it is "fine, fine, fine! just don't
call it UTC!" and that is not economically feasible either IMO.
I don't believe that is his position. I believe his position is more
similar to mine, and that is we have defined UTC to be based on
solar time at the prime meridian, the definition has worked adequately
well for those applications that UTC is intended to handle, and there is
no good reason why we need to change the definition at this time.
If an application cannot handle the leap second's ambiguity, then it
really needs to use TAI.
I mean, honestly, if you need 1 second accuracy in your timing, you
need to use a clock that is precise to 1/10th seconds at the very
least. UTC is really about minutes, not seconds. Ask any metrologist
what happens if he uses a standard that is accurate to the same order
of magnitude as the device he is calibrating.
Leapseconds is not a scientific or astronomical issue, they are
an economical issue.
No, I don't think so. The economy doesn't know what a second is.
-Chuck Harris
In message 42E79591.9010609@erols.com, Chuck Harris writes:
For almost the entire population, time is defined as SI seconds,
60 of which is a minute, 24 of which is an hour, 365 of which is a
year.
That is correct, but you must also remember that 99 and 44/100ths of
the population would do just fine with a timepiece that had no second
hand. Seconds are just a frill to civil timekeeping.
Weeeelll, almost.
You see, the technocratic part of the population is very busy
spinning a technological net around the rest of the population, a
net where seconds can cost you fortunes one example being the mobile
calling-plan where calling out of the time-window cost $100 a minute.
Also, online communities like various games, virtual worlds and
eBay tick on seconds and people get very upset if things don't happen
when they think.
If you have been watching the population in general, you may have noticed
that there is a decline in the number of people that have any interest in how
things work. When I was a kid, there wasn't a toy, or appliance that was
safe from my curiosity. I suspect that you were the same. Now, with the
exception of sadistic impulses that remove the heads from dolls, kids don't
take their toys apart. They just want them to work, and when they stop, they
throw them away.
It may not be their fault. Have you tried taking modern toys apart
? Or a radio ? A TV-set ? There is nothing in there our kids can
learn anything from :-(
The point being, the needs of the population-at-large is not a valid
reason for making a decision as to whether we need leap seconds, or don't.
The population-at-large just doesn't care, at all.
"Doesn't care about leap-seconds" doesn't save your life when some
visual-basic genius has programmed some piece of hi-tech you have
no choice in using or not.
Just because people don't care or notice, doesn't mean not important
to them.
Most people don't care about water, sewers, electricity and civil
order. That doesn't mean it's not important to them. They care
a lot as soon as it doesn't work.
There is a well defined formula that streches indefinitely into the
future, which inserting an extra day almost every fourth year, for
reasons most people do not claim to fully understand.
They don't need to understand it, they just need someone to tell them
that that is the way it is, and they are happy. Don't forget, they don't
care about the details of how things work.
Right, fine. So give me a leap-second formula that works for at
least 25 but preferably 100 years and I'll have no trouble with it.
Certainly we can. As I said earlier, 99 and 44/100ths percent of the
population has no need for seconds whatsoever.
Again, your argument is based on the assumption that just because
people don't now about leap seconds means they won't get harmed
by one. It doesn't work that way.
In fact, they're probably more likely to get in harms way, because
unlike the people who know about the leapsecond and the lack of
proper implementation will probably just wait for the next lift,
just in case.
Robs position, as I understand it is "fine, fine, fine! just don't
call it UTC!" and that is not economically feasible either IMO.
I don't believe that is his position. I believe his position is more
similar to mine, and that is we have defined UTC to be based on
solar time at the prime meridian, the definition has worked adequately
well for those applications that UTC is intended to handle, and there is
no good reason why we need to change the definition at this time.
No, your position is diffrent from Robs, you just don't recognize
the potential for harm at all, Rob at least recognizes that.
Leapseconds is not a scientific or astronomical issue, they are
an economical issue.
No, I don't think so. The economy doesn't know what a second is.
Until after a leap-second hands in the unbudgeted expense or if
we are lucky: the budget request.
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
Hi Poul,
Poul-Henning Kamp wrote:
hand. Seconds are just a frill to civil timekeeping.
Weeeelll, almost.
You see, the technocratic part of the population is very busy
spinning a technological net around the rest of the population, a
net where seconds can cost you fortunes one example being the mobile
calling-plan where calling out of the time-window cost $100 a minute.
Not a problem, those same phones always tell you "network" time. If you
are really that concerned about running out of such a time window, give
yourself a little more slack. Further, telephones generally have a 15 second
aggregation in their billing functions.... a leap-second hardly applies.
Also, online communities like various games, virtual worlds and
eBay tick on seconds and people get very upset if things don't happen
when they think.
I am a devout believer in sniping on ebay. To make my snipes work, I sync
my sniping program to ebay's time. It really doesn't matter to me if ebay is
a little this way, or a little that way of "real" time, as long as I know when they
think the auction will end. (Ebay isn't a good example to argue your cause
because:
a) they have a back channel where they can tell you what time they think it is,
and b) it doesn't matter when you place your bid, only when they receive your bid.
The better examples for your argument would be those where there can be no
regular communication between users of time... nuke submarines for example.)
If you have been watching the population in general, you may have noticed
that there is a decline in the number of people that have any interest in how
things work. When I was a kid, there wasn't a toy, or appliance that was
safe from my curiosity. I suspect that you were the same. Now, with the
exception of sadistic impulses that remove the heads from dolls, kids don't
take their toys apart. They just want them to work, and when they stop, they
throw them away.
It may not be their fault. Have you tried taking modern toys apart
? Or a radio ? A TV-set ? There is nothing in there our kids can
learn anything from :-(
I must be very unusual, I fix modern TV's, radios, and other consumer
electronics doo-dads. It isn't generally economical to do what I do, but
it does keep me in touch with the bleeding edge of consumer manufacturing
techniques.
...
"Doesn't care about leap-seconds" doesn't save your life when some
visual-basic genius has programmed some piece of hi-tech you have
no choice in using or not.
Bad design is bad design. If they didn't botch leap-seconds, it would be
something else. I like open source because I can fix such things if they
bother me. I would never base a product on anything from microsoft.
(I was fired once for relying on microsoft to do their job properly. I try
never to make the same mistake twice.)
Just because people don't care or notice, doesn't mean not important
to them.
Most people don't care about water, sewers, electricity and civil
order. That doesn't mean it's not important to them. They care
a lot as soon as it doesn't work.
Certainly. But what's your point? I don't see these utilities failing
if a second slips here or there. The one case where time is critical
is the power grid, and they keep their own time (Which, IIRC
approximates UTC).
They don't need to understand it, they just need someone to tell them
that that is the way it is, and they are happy. Don't forget, they don't
care about the details of how things work.
Right, fine. So give me a leap-second formula that works for at
least 25 but preferably 100 years and I'll have no trouble with it.
You will have to talk to your favorite deity about that one. We live on
an imperfect planet in an imperfect solar system... I, for one, like that
fact, because the grand scale of the universe's imperfections makes me
feel a little bit better about my own imperfect ways.
Certainly we can. As I said earlier, 99 and 44/100ths percent of the
population has no need for seconds whatsoever.
Again, your argument is based on the assumption that just because
people don't now about leap seconds means they won't get harmed
by one. It doesn't work that way.
In fact, they're probably more likely to get in harms way, because
unlike the people who know about the leapsecond and the lack of
proper implementation will probably just wait for the next lift,
just in case.
I doubt that! The most time knowledgeable function I have ever seen in a lift
is one that sends the idle cars down to the entrance floor during the
morning hours. I hardly think that this will foul up in a dangerous
way if a leap second comes along! You seem to be really stretching to
find some technology that will fail in a way that will vindicate your desire
to kill off leap-seconds.
I don't believe that is his position. I believe his position is more
similar to mine, and that is we have defined UTC to be based on
solar time at the prime meridian, the definition has worked adequately
well for those applications that UTC is intended to handle, and there is
no good reason why we need to change the definition at this time.
No, your position is diffrent from Robs, you just don't recognize
the potential for harm at all, Rob at least recognizes that.
You may think that, but you would be wrong. I see things differently
than you. I don't see a world where the truly critical systems need
to be synced to UTC. Like all of the foibles engineers make, if time is
truly critical to an application, then the application will contain its own
timekeeping, and perhaps its own timescale. (Think NASA and mission
time. ) TAI was developed to handle those cases where seconds needed
to be handled in an unambiguous way.
No, I don't think so. The economy doesn't know what a second is.
Until after a leap-second hands in the unbudgeted expense or if
we are lucky: the budget request.
You won't see any such budget request. None happened 7 years ago,
and none will happen this time either. The time functions where leap seconds
matter will simply get another line added to their tables of anomalies. The
time functions where leap-seconds don't matter (most), will march along as
if nothing happened. The truly critical time functions will continue to use TAI,
or some variant as they currently do.... and folks that attempt to predict future events
down to the second using UTC, will fail because of their lack of forsight...
just as seers have done for time immemorial.
-Chuck
Certainly. But what's your point? I don't see these utilities failing
if a second slips here or there. The one case where time is critical
is the power grid, and they keep their own time (Which, IIRC
approximates UTC).
The long term average of the power grid in the US is 60.000 Hz. Short
term variations from that can and do happen. It is to make sure that
the clocks run on time, on the average. It might be better to say
that the power grid approximates the SI second, since it has no notion
of which second it is (although the control infrastructure for the
grid most likely does).
Warner
The truly critical time functions will continue to use TAI,
or some variant as they currently do....
TAI time isn't a silver bullet. It is a timescale that one can
recover, with some effort, but only if one can get the leapsecond
meta-data from somewhere else, since time is overwhelming distributed
in UTC (even from GPS receivers that could give you TAI, but usually
choose not to). Using it does not eliminate the need to know leap
second information in general, since you have to interface to the
outside world in UTC at some point.
I know this because I've written several control systems that do use
TAI internally, and the issues with leap seconds are large, if you
want to get them right. If you refuse to accept this, then you are
denying reality.
Warner
Certainly. But what's your point? I don't see these utilities failing
if a second slips here or there. The one case where time is critical
is the power grid, and they keep their own time (Which, IIRC
approximates UTC).
The power companies used to use GOES heavily;
early this year the last of them switched to GPS
as a time and phase reference. If there's a power
industry person in time-nuts we'd love to ask you
a few questions.
The long term average of the power grid in the US is 60.000 Hz. Short
term variations from that can and do happen. It is to make sure that
the clocks run on time, on the average. It might be better to say
that the power grid approximates the SI second, since it has no notion
of which second it is (although the control infrastructure for the
grid most likely does).
Warner
For real plots of 60 Hz mains frequency, including
Allan deviation, see:
http://www.leapsecond.com/pages/mains/
The tolerance is good; but so far from the level of
astronomical or atomic seconds that it's not an
issue.
/tvb
Hi Tom:
Does your phase plot mean that a mains powered wall clock might be off
by 10 seconds?
Have Fun,
Brooke Clarke, N6GCE
w/Java http://www.PRC68.com
w/o Java http://www.pacificsites.com/~brooke/PRC68COM.shtml
http://www.precisionclock.com
Tom Van Baak wrote:
. . .
For real plots of 60 Hz mains frequency, including
Allan deviation, see:
. . .
/tvb
I have a digital clock that runs from the 60 hertz power. I have
noticed several times that TVA power can gain time up to 15 seconds
compared to UTC. Takes a few weeks.
I also did a stability test using a rubidium and dividers and I showed
about a +/- 0.03 hertz deviation over one hour, during a 24 hour test
period.
Maybe we all should run a power line test at the same time and we should
be able to see who is leading and lagging !
Brian N4FMN
Brooke Clarke wrote:
Hi Tom:
Does your phase plot mean that a mains powered wall clock might be off
by 10 seconds?
Have Fun,
Brooke Clarke, N6GCE
In message 42E813E4.4090001@erols.com, Chuck Harris writes:
[I split off this topic, it's interesting in its own right I think]
It may not be their fault. Have you tried taking modern toys apart
? Or a radio ? A TV-set ? There is nothing in there our kids can
learn anything from :-(
I must be very unusual, I fix modern TV's, radios, and other consumer
electronics doo-dads. It isn't generally economical to do what I do, but
it does keep me in touch with the bleeding edge of consumer manufacturing
techniques.
Right, but if you were a 7 year old kid, would you learn from it ?
The problem is that microelectronics obscure the basic circuit and
prevents you from poking around with anything but a few peripheral
capacitors which are mostly there for decoupling anyway...
When I took a television apart, there were a schematic pasted on the
back panel, and I could trace the circuit and with a book about
radio reception in hand, I could follow the signals progress. I
could look at the schematic and figure out what happened when I
pushed this button and turned that knob.
If my kid takes a television apart, he can trace any wire with a
signal until it hits an integrated circuit and then what ?
Yes, he'll learn that "it's all the black centipedes which do all
the stuff" and that is both valuable and precise knowledge, but it
is hardly going to make him think electronics is interesting.
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
Does your phase plot mean that a mains powered wall clock might be off
by 10 seconds?
Have Fun,
Brooke Clarke, N6GCE
Yes. I also keep an old AC synchronous motor
wall clock around just to see this effect. To be
fair, it is unusual for it to be off this much, or
for very long.
Maybe we all should run a power line test at the same time and we should
be able to see who is leading and lagging !
Brian N4FMN
You know, if the stackable TAPR module
project catches on another PCB on the list
could be a mains frequency monitor.
It would robustly filter and divide the 50/60 Hz
mains frequency to 1 PPS and then onboard
compare that 1 PPS against the local (OCXO,
atomic, or GPS) 1 PPS to a modest precision
(say, 1 or 10 us).
This would be a cheap board and then many
of us could monitor mains phase and log it in
a standard format. With the right software you
could get plots like this in real-time:
60 Hz AC Mains Frequency Accuracy Measurement
http://www.leapsecond.com/pages/mains/
For extra credit we could follow the GPScon
Z3801A web plot model. See:
International Web Plots
http://www.realhamradio.com/GPS_websites_list.htm
Also, while we're on the subject of monitoring
mains frequency do see Bryan Mumford's page.
Measuring the accuracy of 60 cycle power
http://www.bmumford.com/clocks/60cycle/index.html
If any of you also have an interest in precision
pendulum clocks and don't already have one
of his timers you should look into it:
MicroSet Precision Clock and Watch Timer
http://www.bmumford.com/microset.html
In message 42E813E4.4090001@erols.com, Chuck Harris writes:
Just because people don't care or notice, doesn't mean not important
to them.
Most people don't care about water, sewers, electricity and civil
order. That doesn't mean it's not important to them. They care
a lot as soon as it doesn't work.
Certainly. But what's your point?
It's just above in the first sentence: Just because your neighbor
hasn't heard about leap seconds doesn't mean that they are not
important to him.
No, your position is diffrent from Robs, you just don't recognize
the potential for harm at all, Rob at least recognizes that.
You may think that, but you would be wrong. I see things differently
than you. I don't see a world where the truly critical systems need
to be synced to UTC. Like all of the foibles engineers make, if time is
truly critical to an application, then the application will contain its own
timekeeping, and perhaps its own timescale. (Think NASA and mission
time. ) TAI was developed to handle those cases where seconds needed
to be handled in an unambiguous way.
That's a nice point of view, but experience seems to indicate that you
have not been able to sell it much. Practically anything I see these
days stipulate UTC time.
Until after a leap-second hands in the unbudgeted expense or if
we are lucky: the budget request.
You won't see any such budget request. None happened 7 years ago,
and none will happen this time either.
Have you seriously contemplated the difference between the systems
which were deployed 7 years ago and those deployed today ?
The time functions where leap seconds
matter will simply get another line added to their tables of anomalies. The
time functions where leap-seconds don't matter (most), will march along as
if nothing happened.
And someplaces those two kind of systems meet and trouble ensues...
Because unlike your ideal world, in the real world people slap computer
systems up without thinking about time.
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
In message 20050727.171934.28808808.imp@bsdimp.com, Warner Losh writes:
Certainly. But what's your point? I don't see these utilities failing
if a second slips here or there. The one case where time is critical
is the power grid, and they keep their own time (Which, IIRC
approximates UTC).
The long term average of the power grid in the US is 60.000 Hz.
Are you sure this is still the case ?
Here in the "NordPool" area in nothern europe, 50Hz average went
out the window with the deregulation of the electricity grid because
nobody wanted to be forced to provide the extra power during one
season to gain back what was lost during the other.
And before that, the short term variations were on the order of
a year...
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
In message 002d01c5933c$6c1d2f60$8e18f204@computer, "Tom Van Baak" writes:
You know, if the stackable TAPR module
project catches on another PCB on the list
could be a mains frequency monitor.
It would robustly filter and divide the 50/60 Hz
mains frequency to 1 PPS and then onboard
compare that 1 PPS against the local (OCXO,
atomic, or GPS) 1 PPS to a modest precision
(say, 1 or 10 us).
I actually wrote one in a Z80 many years ago: It would
measure the period of 50 mains cycles (ie: 1 second)
calculate the reciprocal and display frequency with
4 decimals.
Do it on a PIC18F* and feed it a 10MHz atomic clock
and you are done.
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
In message: 87396.1122531897@phk.freebsd.dk
"Poul-Henning Kamp" phk@haven.freebsd.dk writes:
: >> Until after a leap-second hands in the unbudgeted expense or if
: >> we are lucky: the budget request.
: >
: >You won't see any such budget request. None happened 7 years ago,
: >and none will happen this time either.
:
: Have you seriously contemplated the difference between the systems
: which were deployed 7 years ago and those deployed today ?
The Turin leap second survey said that loss of life had occurred due
to a leap second insertion event. The survey didn't provide any
additional details. However, I've not been able to find any media
coverage of such an event, nor find other documentary evidence to
support it.
Warner
The Turin leap second survey said that loss of life had occurred due
to a leap second insertion event. The survey didn't provide any
additional details. However, I've not been able to find any media
coverage of such an event, nor find other documentary evidence to
support it.
Sorry, I haven't ever seen anything on this either.
But on a related note, let's do some math. The
population of Earth is 6.4e9. The birth rate is
21 / 1000 / year so there are 4.3 births per
second (4.3 Hz for the purists ;-). The death
rate is 9 / 1000 / year so there are 1.8 deaths
per second.
That suggests about 4 babies are born and
2 people die during a leap second.
/tvb
TVB said, "If there's a power
industry person in time-nuts we'd love to ask you
a few questions."
I'm not a power industry person, but I've researched the
problem over the years. Here are some of the results.
The generators are all synchronous machines. The ones in
one power plant rotate with very little phase angle
difference between them. Long transmission lines act like
springs, so that phase angles in different plants can be
different, maybe by tens of degrees.
If you had one isolated generator, perhaps the power plant
for a small island, you would have a speed control loop for
the turbine (or engine) throttle that controls power to the
generator shaft. Then you would have a second controller
that could integrate frequency error in a way that trims the
speed control to precisely hold the frequency.
Changes in power delivered to the load by the generator cause
the speed control to change the power delivered to the generator.
If there is a step change in power, like the 5 minute shutdown
of industry at noon on Armistice Day in the 1920s, the power
difference between turbine and generator causes the rotor to
accelerate or decelerate. The speed controller changes the drive
power quickly but not immediately. Stability requires less than
infinite gain. The slower integrating frequency control gradually
brings the frequency error back to zero.
One form of integrating control consists of a person looking at
the powerhouse clock and a standard like WWV. As the error grows,
the person fine-tunes the speed control. Changes may be made a few
times per day.
If the generator is part of a power network then it is not possible
to use integrating frequency control, because you can only have one
of them in a system. Two or more will not track and the result is
wasted power as they fight each other for control. One generator
can't possibly affect the thousands of generators connected to a
power network.
Power distribution networks have dispatchers in regional offices.
As I understand it, the dispatcher watches the time error as well
as the balance of power flows. The dispatcher tells a power plant
how many megawatts to put into the line, not what speed to run.
One of the factors in adjusting the megawatt balance is the time
error.
The US is divided into 3 or 4 independent zones that are isolated
by very high voltage DC transmission lines. Maybe it was 3 - East,
West and Texas.
About 15 years ago, I used a frequency input on a control system to
compare and plot the difference between power line and crystal time.
The power line lagged during the day as industrial loads exceeded
the dispatcher's requests for power and caught up again in the early
morning hours. The swings seldom exceeded +/- six seconds in MN.
I don't know if they do this, but adding a one-second change in UTC
would be no problem at all. Of course, it would take several hours.
Hope this was useful,
Bill Hawkins
At 02:40 AM 7/28/2005, M. Warner Losh wrote...
The Turin leap second survey said that loss of life had occurred due
to a leap second insertion event.
That is a deliberately misleading statement. It MUST be the case that the loss of life occurred due to and improperly designed, incorrectly specified, or improperly used system. The person/organization at fault seeks to misplace blame.
Tom Van Baak wrote:
You know, if the stackable TAPR module
project catches on another PCB on the list
could be a mains frequency monitor.
It would robustly filter and divide the 50/60 Hz
mains frequency to 1 PPS and then onboard
compare that 1 PPS against the local (OCXO,
atomic, or GPS) 1 PPS to a modest precision
(say, 1 or 10 us).
This would be a cheap board and then many
of us could monitor mains phase and log it in
a standard format. With the right software you
could get plots like this in real-time:
I'm all for it -- maybe not as a formal TAPR project (because I suspect
the volume would be fairly low) but certainly we could share a design,
and I strongly suspect (but can't promise) that TAPR could at least make
the boards available, if not a full kit.
Someone want to take on the design challenge? If so, contact me
off-list and I can provide the mechanical and electrical details for the
stacked board design -- basically, it's just a standard board size with
standardized locations for up to 6 output BNCs, one input BNC, and
headers so the raw DC in and reference in can be routed to boards above
and below.
John
Mike S wrote:
At 02:40 AM 7/28/2005, M. Warner Losh wrote...
The Turin leap second survey said that loss of life had occurred due
to a leap second insertion event.
That is a deliberately misleading statement. It MUST be the case that the loss of life occurred due to and improperly designed, incorrectly specified, or improperly used system. The person/organization at fault seeks to misplace blame.
I've stayed out of this argument, because frankly I don't have a strong
view on the subject, but I don't think this is misleading at all. The
point Warner and Poul-Henning have been trying to make is that leap
seconds will cause programming errors, and this seems to be (anecdotal
and undetailed) evidence of that.
Certainly the death (if it occurred) was not an automatic result of the
leapsecond, but rather was the result of something that broke because it
wasn't properly programmed to deal with the leapsecond. But that's the
one of the points the anti-leapsecond folks are trying to make -- things
will break, because the majority of programmers don't know how to, or
that they even need to, program for leapseconds, and the infrequent and
unpredictable occurrence of leapseconds makes it unlikely that situation
will change.
John
In message: 6.2.3.4.0.20050728064100.02d9dc68@mjs.alientech.net
mikes@flatsurface.com (Mike S) writes:
: At 02:40 AM 7/28/2005, M. Warner Losh wrote...
:
: >The Turin leap second survey said that loss of life had occurred due
: >to a leap second insertion event.
: That is a deliberately misleading statement. It MUST be the case
: that the loss of life occurred due to and improperly designed,
: incorrectly specified, or improperly used system. The
: person/organization at fault seeks to misplace blame.
Well duh! Designing things is hard. Let's go shopping instead! The
point is that there's a number of real world consequences to badly
designed thing. That's part of the cost of getting something right,
isn't it?
Warner
At 08:29 AM 7/28/2005, John Ackermann N8UR wrote...
Mike S wrote:
That is a deliberately misleading statement. It MUST be the case that the loss of life occurred due to and improperly designed, incorrectly specified, or improperly used system. The person/organization at fault seeks to misplace blame.
that leap seconds will cause programming errors,
Programmers cause programming errors. Leap seconds may make them apparent.
Certainly the death (if it occurred) was not an automatic result of the leapsecond, but rather was the result of something that broke because it wasn't properly programmed to deal with the leapsecond.
The counter argument is that removing leapseconds will break properly implemented systems in unknown ways, the blame will them be not with someone who did things in violation of a well documented specification, but with those who changed the specification in a fundamentally incompatible way for selfish reasons.
mikes@flatsurface.com (Mike S) writes:
Programmers cause programming errors. Leap seconds may make them apparent.
Certainly the death (if it occurred) was not an automatic result of the leapsecond, but rather was the result of something that broke because it wasn't properly programmed to deal with the leapsecond.
The counter argument is that removing leapseconds will break properly implemented systems in unknown ways, the blame will them be not with someone who did things in violation of a well documented specification, but with those who changed the specification in a fundamentally incompatible way for selfish reasons.
How does a properly implemented system accounting for leapseconds fail
when leapseconds fail to come? Sure there will be unnessesary code
that could be removed. But I do not see why the system would break.
--
Björn
In message m3k6jabyoh.fsf@lysator.liu.se, Bjorn Gabrielsson writes:
mikes@flatsurface.com (Mike S) writes:
Programmers cause programming errors. Leap seconds may make them apparent.
=20
Certainly the death (if it occurred) was not an automatic result of the =
leapsecond, but rather was the result of something that broke because it wa=
sn't properly programmed to deal with the leapsecond.
=20
The counter argument is that removing leapseconds will break properly imp=
lemented systems in unknown ways, the blame will them be not with someone w=
ho did things in violation of a well documented specification, but with tho=
se who changed the specification in a fundamentally incompatible way for se=
lfish reasons.
How does a properly implemented system accounting for leapseconds fail
when leapseconds fail to come? Sure there will be unnessesary code
that could be removed. But I do not see why the system would break.
My interpretation of this is that systems which assume that DUT < 1s
fail, when leap seconds are applied.
That's probably true, but since DUT is only relevant if you study
extraterrestial objects, we can safely assume that 99.9% or more
of those systems involve astronomers and optics.
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.