On 11/22/21 10:36 AM, j...@trillian.mit.edu wrote:
Some time back, I saw a fun discussion started with one person
arguing for year-round EDT in New Enland, followed by someone else
suggesting that instead, New England should switch to AST. The fun
part was the quick realization that a lot of people didn't understand
the difference, and got into rather nasty fights calling each other
all sorts of insulting names.
Since then I've often thought that this is a really good way to mess
up such discussions, and we should all be trying it whenever the
topic comes up. ;-)
I love it! Ask those year-arounders whether they want EDT or AST, and
then argue for the opposite.
Time is real simple. And very useful for predicting and scheduling and
synchronizing stuff.
The very useful aspects mean we layer lots extra important stuff on top
of it: where the sun will be in the sky, seasons, birthdays, religious
holidays (haven't wars been fought over when is Easter?), precise
durations, predictable decompositions, etc. And let's also change our
clocks twice a year, in most but not all places, and let's do it on
dates which are not consistent from one participating place to place nor
from year to year. Also allow local officials to change timezone for
arbitrary reasons, changing their minds form time to time, including
fractional timezones!
Which means time is no longer simple. But multi-year mortgage and bond
calculations still need to be correct.
At first glance time still looks simple, and that's when people start
writing bugs.
When I was born the second was defined as a specific fraction of a day,
which meant the length of a second was variable depending on what the
earth was up to. When I was a little kid the second was redefined to be
of a fixed length, and it was the *day* that became of variable length;
the second was heretofore defined as rock-solid.
Nope. Youngsters know time is simple and still write code that assumes
things are today as they were when I was born: a second is a fixed
fraction of a day.
So because programmers think time is simple, we have a fairly new
(fuzzy) Google definition of the second. Mostly it is dang precise and
stable, but every year or so, it starts to slew wildly away from its
usual precise duration and then slew wildly back, all to avoid leap
seconds that would trigger bugs where programmers thought they knew that
a minute would always have 60-seconds. (Silly of them.)
I say fuzzy because I am pretty sure how and when the slewing happens is
not well defined, is probably not consistent from one leap second to the
next. And this odd time standard is distributed via NTP, which was not
intended to distribute a non-stable reference, so the result is going to
be a mess from any time-standardization perspective. But at least it
won't trigger any brand new 60-second-per-minute bugs.
And god help any programs that crazily assume a second is of fixed
length, but then get hooked up to Google time.
Time is complicated.
-kb
_______________________________________________
Discuss mailing list
Discuss@lists.blu.org
http://lists.blu.org/mailman/listinfo/discuss
