[LEAPSECS] Schedule for success

[Poul-Henning Kamp]

In message <495AEA36.7080608 at cox.net>, Greg Hennessy writes:


>Poul-Henning Kamp wrote:

>> In message <495AC9C8.9010601 at cox.net>, Greg Hennessy writes:

>> 

>>> If you wish to make an argument that they did get it incorrect, please

>>> do so.

>> 

>> GMT is a timescale owned and defined by the British Parliament.

>> 

>> I rest my case.

>

>Excuse me, I don't see a case.

>

>You seem to be asserting that the use of GMT is an

>error. What is the error that you see?


Greg,

Several people on the "pro-leap" side here, have argued that
"programmers should just do it correctly and we would have no
problems. 

I don't know what you do for a living, but unless you are a programmer
yourself, you are likely to work at a level of complexity about
1000 times lower than programmers like Warner and me.

Right now, despite the best of my efforts, my "Varnish" program
which is used to speed up websites like marines.com, is growing
larger than the on-board software in the apollo capsule.

And that is widely admired for being an incredibly small program
for what it does.  The nearest competitor is five times larger
(and ten times slower :-).

The FreeBSD kernel Warner and I have worked on for 15 years approaches
5 million lines of code.

For comparison, the US navy touted the complexity of the Air
craft carrier Ronald Reagan (CVN-76), by boasting that it was
built from "more than one million individually tracked pieces".

Spare parts for al US cars, numbers only approx 1.5 million.

Of the 200+ FreeBSD developers, only Warner and I truly know how
the relevant 2500 lines of code keep track of time and leap-seconds.

Me because I wrote most of it, Warner because he worked at timing.com
who used it to deliver timing gadgets to customers like NIST and
USCG.

I won't claim. that Warner and I are the best programmers there
are, when it comes to timekeeping in operating system kernels.

But you would be hard pressed to find ten like us.  Anywhere.

Therefore I do feel on solid ground when I tell you, that when not
even NASA can get a detail like UTC/GMT right, we have no reason
to expect or rely on programmers to get leap seconds right.

Emperical evidence show that this is indeed the case, and the
fact that even national time laboratories screw up, underscores
this fact: http://phk.freebsd.dk/Leap/20051231_HBG/

Back when computers had just broken the 32 kilowords barrier, one
of the greatest thinkers in computer science wrote:

     "My own feelings are perhaps best described by saying that I am
      perfectly aware that there is no Royal Road to Mathematics, in other
      words, that I have only a very small head and must live with it."
		-- Edsger W. Dijkstra

We have no way to upgrade our programmers, we can only make the
task in front of them simpler.

The fact that you cannot see the problem, does not make it a non-problem,
people like Warner and me have seen the problems, we know they are there.

If you don't trust us, how about this:

	Judah Levine of the National Institute of Standards and
	Technology in Boulder, Colorado, which provides the time
	standard and technical support for most commercial activities
	in the US is braced for New Year's Eve. "On December 31,
	I'll be waiting with a cup of coffee for the problems to
	roll in," he sighs.

	http://www.newscientist.com/article/mg20026875.400-calls-to-scrap-the-leap-second-grow.html


Poul-Henning

-- 
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.