On Tue, 20 Jul 2004, Eugene van der Pijll wrote:

> Yes, but the difference on 1971-12-31 is undefined. For our utc time zone,
> we could define "utc"-TAI = 9s on 1971-12-31. Why we would want do do that,
> I don't know. But we could.

Leap-seconds are officially declared by the IERS but whatever.  The current 
implementation makes the delta effectively 9s on 1972-01-01, which is *wrong*.

> That's the one I meant. I don't think there's much user code out there that
> even cares about leap seconds.

Everyones code cares about leap-seconds depending on how you look at it.  Imagine 
someone writing an "epoch" timestamp, upgrading DateTime, and reading in the old 
timestamp.  There is now a 1 second error - Ouch.  The same thing applies for someone 
with a broken version of DateTime reading in a timestamp that was generated with a 
fixed version.  Or what about the people parsing stamps not generated by DateTime?

Cheers,

-J

--

Reply via email to