Re: [Ntp] Leap second draft

[Watson Ladd <[e-mail]>]

On Thu, Jan 23, 2020, 10:51 PM Patrik Fältström <[e-mail]> wrote:

> On 24 Jan 2020, at 0:36, Warner Losh wrote:
>
> >> POSIX doesn't specify, and so systems choose a number of different ways
> of coping (repeat the last second of the day, repeat the first second of
> the next day, freeze time during leap second, do some crazy smear to
> pretend reality matches the POSIX spec, kernel panic, etc).
> >
> >
> >
> > This part is true. It's the biggest flaw with time_t: It doesn't
> acknowledge that leap seconds exist, so what to do during that thing which
> the standard says doesn't exist is up for grabs :(
>
> As the POSIX definition of "number of seconds since the epoch" is "number
> of days since the epoch plus number of seconds within the current day" the
> last second of one day and the first on the next have the same number.
>
> The fix is of course to be made at POSIX and if it was done, tons of
> issues would just go poof as time then always did increase.
>

Why does POSIX matter for NTP timestamps? We can always clarify the
language independently.


> There would still be an issue with "what time is it a gazillion seconds
> from now" as one do not know the number of leap seconds to be added in the
> future.
>
> But lets not mix up the issues.
>
> Anyway, I have seen proposals in POSIX on fixing this which have not been
> moved forward. Anyone being involved in POSIX that knows anything about
> that status?
>
>    Patrik
>