[LEAPSECS] A standard for leap second smearing

Steve Summit scs+ls at eskimo.com
Wed Sep 28 09:23:20 EDT 2016


Richard Thomas wrote:
> On Sep 27, 2016, at 7:41 PM, Steve Summit <scs+ls at eskimo.com> wrote:
> > It just feels right to me that any shenanigans should be confined
> > to the day that the leap second is at the end of, and that on the
> > next day, everything is back to normal.
>
> Remember that the leap second occurs at midnight UTC, not midnight your local
> timezone.

Indeed.

> Does this change anything?

If a leap second manifests as an awkward 1s jump in Posix time,
it could matter a lot.  But if we're smearing the leap so that
it's invisible to all but the most discerning of clients, I don't
believe it matters when it happens.

> > An alternative is to do the smearing purely mathematically,
> > between (2) and (3), notionally as times are being handed to
> > user mode by gettimeofday() and the like.  I don't believe it's
> > necessary to tinker with internal clock steps or frequencies.
>
> This is far and away the easiest solution.
> It only involves code changes in one place.
> And we already have a good idea of what the code should look like,
> because we have the 'right' timezone family that converts between
> UTC and TAI.

I'd say that's different, though, because the 'right' code *does*
make leap seconds explicitly visible; it does *not* smear them away.

> In fact, I'd go even further and suggest that the standard should require
> that (1) and (2) should be TAI, not UTC.

Hoo, boy.  Me, I'd say that's a *whole* different kettle of fish.


More information about the LEAPSECS mailing list