[fitsbits] Re: leap second alert

[William Thompson]

pausch at merope.saaf.se (Paul Schlyter) writes:

>In article <838ofn$7v at post.gsfc.nasa.gov>,
>William Thompson <thompson at orpheus.nascom.nasa.gov> wrote:

	(stuff deleted)

>> Personally, I would expect UT to mean either UTC or UT1.  This is one reason
>> why letting UTC drift relative to UT1 is so troubling.
> 
>If one requires an accuracy large enough that the difference between
>UT1 and UTC matters, then one should not just write "UT" (or "GMT"),
>but instead explicitly specify if UTC or UT1 is used.

Well, right now the difference between UT1 and UTC is guaranteed to be less
than a second, but the proposal under discussion would change that
significantly.  If the proposal were adopted, they'd drift by about a minute
per century!

>> The computer industry tends to use the acronym GMT as a synonym for UTC, but
>> there's currently no exact recognized definition for GMT.
> 
>There have been several different definitions of GMT.  Earlier, GMT was
>often considered equal to UT1.  Thus, the term "GMT" should be used only
>when the difference between UTC and UT1 doesn't matter.
> 
>> If they're so
>> worried about handling leap seconds in their software, they can define
>> 
>>         GMT = TAI - 32
>>
>> (or whatever the value of TAI-UTC is when they freeze it).  The values
>> of GMT and UTC would start out in sync, but then would start to diverge
>> as soon as the next leap second was implemented in UTC.
> 
>No they shouldn't -- there are enough different definitions for GMT already,
>no need to invent yet another one!
> 
>If one needs to use a uniform time scale with no leap seconds, there are
>already two such time scales available:  TT and TAI.  Choose one of those!

Yeah, with hindsight I realize that's a silly idea.  Consider it withdrawn.

William Thompson