[LEAPSECS] Look Before You Leap ? The Coming Leap Second and AWS | Hacker News

[Eric R. Smith]

On 19/05/15 08:30 PM, Joseph M Gwinn wrote:
>> From: "Eric R. Smith" <ersmith at hfx.eastlink.ca>
>> To: Leap Second Discussion List <leapsecs at leapsecond.com>
>>> True UTC (with leap seconds) didn't cure a problem the committee cared
>>> about, and managed to cause problems they did care about.  In short,
> POSIX
>>> systems have to be able to work in a cave, with no access to the sky or
>>> knowledge of astronomy.
>>
>> If POSIX time_t were actually a count of SI seconds elapsed since the
>> epoch, then a machine in a cave (with an accurate enough clock) could in
>> principle maintain correct timestamps. As it stands though, POSIX time_t
>> cannot be implemented without access to a UTC reference of some kind,
>> i.e. access to the sky.
> 
> Well, while POSIX mentions SI seconds, the standard is careful to say that
> these seconds are not exactly SI seconds (because UNIX workstations can
> have pretty bad clocks).  And the standard specifically disclaims being
> UTC, despite the appearance.  Read the standard carefully.  It is intended
> and designed to support isolated operation.

I don't have the actual standard in front of me, but have seen claims
that POSIX time_t is defined (for years after 1970) to be:

tm_sec + tm_min*60 + tm_hour*3600 + tm_yday*86400 +
    (tm_year-70)*31536000 + ((tm_year-69)/4)*86400 -
    ((tm_year-1)/100)*86400 + ((tm_year+299)/400)*86400

and that "each and every day shall be accounted for by exactly 86400
seconds". Is this correct? Since the length of the day is not in fact
exactly 86400 SI seconds, it would follow that a POSIX compliant system
has to know how many days have elapsed since the epoch, i.e. it needs to
have some kind of access to the sky. Am I misunderstanding something?

Regards,
Eric