On Sat, 28 Jun 2014 09:53:58 -0400, John Gilmore wrote:
>
>Implicit in Chuck's original post is that he also has a date
>available. He needs a 4tuple:
>
>o date,
>
>o time,
>
>o location, specified as a latitude-longitude pair, and
>
>o identity, UTC|xDT
>
True.
>...
>In general, date and time values are inseparable. The international
>date line exists and cannot be wished away. You will need do your
>conversions on a date-time value, not on a time value alone.
>
True. Consider the alternative, that the date should change simultaneously
world-wide, but not everywhere at midnight. (A more palatable alternative
is world-wide use of UTC.) Star Date?
>... Again, time-zone UTC offsets are not in
>general integers. The Australian Cocos Islands Time Zone is UTC+6.30.
>
But that's an integral number of minutes. I know of no offset that's
not an integral number of minutes, but it may not always have been so:
http://en.wikipedia.org/wiki/Standard_time#North_America
>Skip Robinson is clearly right. A table-driven solution is ineluctable.
>
True.
>You can get help, crucial help, from the IANA database, but in my
>experience it is never usable directly. You will need to extract the
>relevant data it contains, putting them into your own tables, which
>you will need to regenerate in at least quasi-automated fashion every
>time the IANA database is updated.
>
Most UNIX(-like) systems provide the service.
>It is common to grossly underestimate the complexity of date-time
>calculations, probably because they are usually botched without
>recognizing that they are being botched.
>
To wit, z/OS.
History questions (I believe UTC was introduced in 1972):
When was the TOD clock introduced?
When was the Sysplex Timer (AKA ETR) introduced?
When was CVTLSO introduced?
I believe that the TOD clock runs 10 seconds behind TAI (in a sense)
in order than no discontinuity in TOD occur in 1972 when the reference
switched from Earth rotation to an atomic standard. But was this a
great concern for MVS(?) in 1972?
-- gil
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
