[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: leap second



mbaker@iee.org (Mark Baker)  wrote on 21.06.97 in <E0wfLXG-0008VO-00@mnb20.pet.cam.ac.uk>:

> In article <6ZDgv9zzcsB@khms.westfalen.de>,
> 	kai@khms.westfalen.de (Kai Henningsen) writes:
>
> > Consider a system using "real" time. On June 31, its idea of time would be
> > wrong until the next software upgrade.
>
> No. Using real time, the system clock increments normally, and correctly
> measures the time since the epoch. The conversion from system time to local
> time changes every time there is a leap second.

Exactly. But how does the system know about leap seconds? Only by  
upgrading the software.

> > Then, all time stamps would suddenly change by one second (possibly
> > causing FTP server remirroring and other unpleasant effects).
>
> No, because they are based on the system time which is consistent. This
> would be a good thing.

This turns out not to be the case.

You know, I have a private Debian mirror. I know whereof I speak. FTP and  
mirror usually both use local time, not GMT. Changing the GMT-local  
conversion _does_ cause remirroring. I've seen it happen.

> As it is, we use POSIX time, which means that the system time follows GMT.
> When there is a leap second the time itself is changed; the timezone
> information does not need to.

And this is a good thing. (It's also completely automatic if you use  
xntpd.)

> > This is completely unacceptable. OS time must be predictable.
>
> Which is why real time would be much better than POSIX time.

I cannot rightfully understand the mental confusion that leads to  
sentences like this. (Slightly modified from a Babbage quote about GIGO, I  
believe.)

POSIX time is completely predictable.
"Real" or "right" time is completely unpredictable.

> Unfortunately we have to use POSIX time, so we're compatible with other
> computers on the network :(

Fortunately, we can use POSIX time, which is the only sane time for  
computers.

There's one way of doing it that can work completely automatic, no user  
intervention.

There's another way of doing it that needs software updates whenever a  
leap second is introduced, and which then invalidates local time for any  
time after that leap second insertion.

Why on earth would any rational person want to use the latter, except for  
some very few specialized applications?

MfG Kai


--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-devel-request@lists.debian.org . 
Trouble?  e-mail to templin@bucknell.edu .


Reply to: