[ntp:questions] A Suggestion For Abolishing the Leap Second

Rob van der Putten rob at sput.nl
Wed Jun 6 10:47:52 UTC 2007

Hi there

Casper H.S. Dik wrote:

> But it barely works as witnessed when the US changed the rules
> yet again; it caused Californian off-peak rates to shift an
> hour because the meters could not deal with the change
> (so the off-peak rate was charged for some of peak and vice
> versa)

Systems should use some sort of epoch internally. Anything else is 
broken by design.
Time zone info is just there for the convenience of human readers, and 
not something to be used by machines.

> Because the on-disk timestamp format cannot be retroactively changed.

OS-es use some sort of epoch. It doesn't matter whether this is UTC or TAI.
For instance, all my tables in my MySQL server have a timestamp. On 
queries is returns something like 20070606120233. Internally it stores 
the time as seconds since Thu Jan  1 00:00:00 1970 UTC. And the message 
id in my previous post starts with 46658D83, which is the hex epoch.
If my system was using TAI instead of UTC, it would have used 46658DA4 
instead, which is still a valid message id.

I can hardly imagine a file timestamp not using epoch.

Nothing is more surreal then reality

More information about the questions mailing list