Docket No. 2005-22114

[Paul Eggert]

"David Braverman" <david at inner-drive.com> writes:

> isn't it acceptable to allow computers to make two time changes that
> night without worrying about the human impact?

I would say not.

> For example, radio stations I have worked for in the past almost always
> switched the logs to or from DST when the next engineer came on.

But eventually, as those radio stations and other clocks get
automated, and as people's clocks reflect what the tz data actually
say, we will see problems if the times are not what people expect.  I
suspect and hope that the problems will not be catastrophic, but it's
conceivable that they will be, and our best course is to make them
less likely by having the clocks reflect common practice.

This may help to explain why the tz database attempts to reflect the
clocks that people actually see in consensus practice, even when this
is not the "official" time.  We can't always achieve that goal, of
course (partly because the actual behavior is not that well
documented), but in this particular case it's quite clear that the
official rules disagree with what will actually happen.

> I'm just not sure why the debate has continued so long, when the
> regulation is as clear as any I've seen, even though it has a silly
> unintended effect (which is often the hallmark of regulations in
> general).

Well, if we weren't so obsessive about doing time "correctly" we
probably wouldn't be on this mailing list....

This situation reminds of Brazil's Presidential Decree 4,844
(2003-09-24), which had a somewhat-similar bug before it was corrected
the next day; see the red print at the bottom of
<http://www.presidencia.gov.br/CCIVIL/decreto/2003/D4844.htm>.

It is a bit ironic that the US is less efficient than Brazil in fixing
obviously-broken regulations like this.