Re: localtime()

On Mon, 5 Dec 2011, Peter Conrad wrote:

> > >    Therefore I think your conclusion that counting
> > >    leap seconds is required by law is wrong.
> > 
> > It’s implicit (and more verbose in ZeitG): to be able
> > to have a Jan 1 00:59:60 CET, the leap second must
> > exist in the system.
> 
> yes and no. It's a question of representation. The law does not
> require the system to have a unix timestamp matching Jan 1 00:59:60 CET.

I’d say it does, but let’s leave that to the sharks, erm lawyers.
I guess nobody will sue you over not doing it, but still, it’s a
nice point (especially when having breakfast with Plan 9’s Uriel)
if you can point out POSIX is probably illegal in Germany ;-)

> You *can* have applications which are able to process such a time
> value without having a matching unix timestamp. I think database

Yes. We’re not talking about applications here but POSIX.

> You can also run something like ntpd which lets your system clock

That would be off a few times over the lifetime of a system.
Also, it requires “manual” intervention to keep a correct
time which I don’t like but others (like FreeBSD’s phk) may
agree to live with.

bye,
//mirabilos
-- 
15:41⎜<Lo-lan-do:#fusionforge> Somebody write a testsuite for helloworld :-)