BSD sleep

Matthew Seaman matthew at
Wed May 29 06:13:09 UTC 2013

On 29/05/2013 05:59, Michael Sierchio wrote:
> On Tue, May 28, 2013 at 4:45 PM, Joshua Isom <jrisom at> wrote:
>> You think it's trivial until you read this:
>> programmers-believe-about-time<>
> Some days have 86400 seconds, some have 86401.  There is a provision for
> two leap seconds to be applied at once, but that hasn't ever happened.
>  Still, a truly correct clock, set to UTC, might someday read
> 23:59:59
> 23:59:60
> 23:59:61
> 00:00:00
> How many seconds did that hour have?

Right.  The fact that on very rare occasions a minute may not have 60
seconds in it plus many other corner cases in calculating the current
wall-clock time is an amusing irrelevance.

First of all, sleep deals in local elapsed time, which is a well defined
property even if the displayed wall-clock time would be all over the
place due to DST changes or relativistic effects or whatever.

In this case, I'd be pretty surprised if GNU sleep's algorithm was
anything more complicated than to convert the stated time into seconds
and then sleep that number of seconds.  And to do that conversion, it
wwould just define one minute as 60 seconds, one hour as 60 minutes, one
day as 24 hours, one week as 7 days, perhaps one month as 30 days, one
year as 365 days[*].  Sure, it's simplistic and unsophisticated, but as
an engineering solution it's good enough for the vast majority of



[*] I haven't checked on GNU sleep, but (for example) this is exactly
what dnssec-keygen(8) does.

Dr Matthew J Seaman MA, D.Phil.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 268 bytes
Desc: OpenPGP digital signature
URL: <>

More information about the freebsd-questions mailing list