[time-nuts] Re: UTC - A Cautionary Tale

Poul-Henning Kamp phk at phk.freebsd.dk
Thu Jul 14 16:48:01 EDT 2005


In message <BB586645-2116-4EC5-A87D-2A2066C16D11 at noao.edu>, Rob Seaman writes:

>But note that this includes millions of amateur astronomers as well.   
>If you haven't looked recently at commercially available amateur  
>telescopes, they are paragons of civil time handling.

I know.  I upgraded my old 4" Newtonian to an ETX125 last year,
despite the fact that there are only 50 decent nights a year in
Denmark and they're all Tuesdays :-(

The ones I've seen are programmed to tolerate up to one minute
of wrong time because people tend to use a rather random wrist
watch to set the time, and the internal RTC is not much better.

One minute UT1-UTC difference is probably 50 years or so, and all
things considered, I doubt the telescopes will last that long.

Besides all of these scopes can update their software with a PC.

And knowing amateur astronomers, having to look up the UT1-UTC
difference once per year would probably only seem even more geeky
to them :-)

For the high end scopes with built in GPS receiver, the US has
already indicated that the UT1-UTC difference could be added to
the almanac of the GPS signal.

>Shouldn't we explore the requirements and use cases before making a  
>change to the standard?

Absolutely, but shouldn't we look at more than astronomy while
doing so ?

I will agree with you that the current process for decision is
somewhat flawed seen from a democratic point of view, but on
the other hand, leap seconds is not something you can decide
in a large group of people, Parkinsons Law will simply prevent
a decision.  (for more details see:)

http://www.freebsd.org/doc/en_US.ISO8859-1/books/faq/misc.html#BIKESHED-PAINTING

>The crystal clear intent is simply to cease issuing leap seconds.   

Yes, and after having considered things carefully, I fully agree
with both the sneaky approach (see above) and the desire to get
leap seconds killed dead.  Soon.

The irony of fact that we seem to have gotten one last leapsecond
may in my opinion be just what it takes to get them killed, and if
so, the potential havoc and expense of the next leapsecond is well
spent (again: IMO).

The matter of the fact is that we have four groups of people here:

	1) Astronomers
		< 1 million

	2) Programmers who do leapseconds.
		< 1000

	3) Programmers who don't know what leapseconds are.
		> 2 million

	4) The rest of the world.
		> 6 billion.

Only Group 1) wants leapseconds because they are used to having them.

Only Group 2) wants to get rid of leapseconds, and their argument is
that leapseconds means that Group 3) makes a mess for group 4).

As far as I can tell, somebody in group 4) realized what it would
cost to upgrade group 3) people to group 2), and deciced that group
1) can take a hike and find another solution.

>Many systems and many users for many purposes certainly  
>won't care.  Shouldn't we try to identify ones (oh - air traffic  
>control springs to mind) that might *possibly* care (if so, big time)  
>and perform a Y2K-like inventory of ramifications?  I am skeptical  
>that projects that have inappropriately selected UTC as a timebase  
>and have inappropriately implemented support for what is a  
>straightforward standard will make any better choices in the future  
>or will ever implement systems that better conform to any new  
>standard.  Ostriches don't actually put their heads in the sand - nor  
>should we.

The problem is one of magnitude.

Just carrying out the "Y2K-like inventory" will cost more money
than the astronomers would ever dream of getting their hands on.

Rather than see the Athur Andersons of the world make tons of money
to write reports which say "Something might, or might not, happen."
like they did for Y2K, I'd far rather see the astronomers say "Well,
Yeah, we can see this is inconvenient for the rest of you and we
can do this change for $N".

And I'd only be more happy if the astronomers manage to slip a
couple of new observation facilities in while they're at it.

To put things in perspective:  One civil danish regulatory entity
spent more money on Y2K audits than gets spent on astronomy in
Denmark in a year.

I have no idea how much money the military spent, how much money
the other civil regulatory entities spent, and how much money private
companies spent, but I'll tell you this:  You could get a lot of
astronomy for that pile of money.

Not that you will, mind you, I know politicians as well as you do :-(

>This has never been a discussion about whether TAI is "better" than  
>UTC - for some purposes, of course it is.

The problem is that UTC is printed in too many documents.  The cost
of having para-legals track it all down and passing ammendments to
change it all to TAI or whatever we decide to call it would cost
a fortune you cannot even dream about. (And thats not even counting
all the pork that would be attached to such bills).

And then comes the issue of explaining to everybody how to convert
"from old to new time" etc.

By simply dropping leap seconds from UTC, the group 3) and group 4)
above won't notice a thing and we saves billions of dollars.

I know it's not scientifically optimal and I know we'll all
get blindly drunk and moan "my government doesn't understand me"
but the other thing is just not going to happen, so we might
as well get on with it and get it over with.

>Civil time should continue to reflect time-of-day.  The  
>headaches from retiring leap seconds will prove larger than those  
>from continuing to issue leap seconds.

This is where I just don't think you grasp the scale of things
in the non-astronomy part of the world.

>Astronomers are among the few communities whose line of work is  
>absolutely guaranteed to be interrupted by a leap second occurring at  
>midnight on New Year's Eve.

Cry me a river...

So far I already know about two medical factories, three automated
distribution warehouses, one air traffic control zone and a number
of more usual IT installations which have already declared major
disruption to their production plans on new years eve.

And I live in Denmark where most stuff is shut down (firmly!) for
new years eve.  Ask in Asia if they care...

>Observations throughout the western hemisphere will pause before  
>midnight and resume afterwards.  Others will simply continue with the  
>thought that the effects of the leap second (if any) will calibrate  
>out during reductions.  We tolerate these complications because they  
>are the price to pay for retaining sanity in our clocks.

And you think the rest of the world think your clocks is a good
enough argument that we have to pay the bill for down time, extra
man-power and all that stuff ?

If you can sell that can sell anything...

>I perused a number of the back threads from this list.  A recurring  
>topic appears to be the complexity of systems that are used to  
>transport time.  The problem is not that the POSIX (or what have you)  
>time model is too complex.  The problem is that it is too simple by  
>far.

Trust me, changing UTC is a picnic compared to changing POSIX time_t
definition.

We're not just talking about the group 3) people here, we're talking
about the marketing departments of the entire OS business, Microsoft,
IBM, Apple, Sun, Linux, FreeBSD etc.  They couldn't ever agree what
color fire should have if their life depended on it.

>This is a task worthy of our best efforts.

No, this is a windmill, and the effect of challenging it to duel is
described in detail in a book by Cervantes.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at 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.




More information about the time-nuts mailing list