[ntp:hackers] ITU and Leap second elimination

juergen perlinger juergen.perlinger at t-online.de
Thu Oct 3 09:45:58 UTC 2013


On 10/03/2013 11:07 AM, Poul-Henning Kamp wrote:
> In message <524D30CD.5010309 at t-online.de>, juergen perlinger writes:
>
>> 'Put it in the application' would be the most
>> sensible approach, could we only start on green fields.
> I don't agree.
>
> Application programmers are the least competent group of programmers
> we could possibly ask to deal with this.
>
> The right approach would be to have the system offer APIs which
> makes it very hard, preferably impossible, for the applications to
> get it wrong.
>
> However, that boat has long since sailed...
>
I was too terse, probably. What I meant was the distinction kernel vs.
application. Then user space libraries could have taken over. Only few
people deal with the intricated math of calendars and time zones, but
'localtime()' is used a lot. Not always properly, I have to admit. There
was a reason why I wrote a calendar module for NTPD.

About fool-proof APIs, I've come to side along with Murphy's Law. No
matter what safeguards you install, someone will find a new creative way
to make a mess. I agree that it should be as hard as possible, but after
25 years of software development I have abandoned hope in that
particular spot.


More information about the hackers mailing list