[tz] 2013g - Morocco

Stephen Colebourne scolebourne at joda.org
Fri Oct 4 10:24:28 UTC 2013


On 4 October 2013 10:18, Ian Abbott <abbotti at mev.co.uk> wrote:
> On 2013-10-03 21:11, Paul Eggert wrote:
>>
>> This predicts that Moroccans will change their clocks twice in 8 days.
>> Not as weird, but still pretty weird, and it's tempting to omit the
>> first two transitions.  But then the question is, where does one stop
>> omitting transitions?
>>
>> The answer to this question depends on how much of a stickler
>> Moroccans will be for following the rules, and that involves complex
>> religious-commercial-political tradeoffs that I don't pretend to
>> understand.  In the current experimental version I didn't omit any of
>> the predicted transitions, as that was one less thing for me to worry
>> about, but if there's consensus to omit them if they would imply
>> moving the clocks less than N days apart we can do that too.
>
>
> IMHO it's not worth worrying about.  It's not as if any of the listed
> transition dates (particularly the "suspension of DST during Ramadan" dates)
> are fixed in stone anyway.  The Moroccan government may decide to suspend
> DST a day earlier than predicted one year for example.

Exactly. The future rules are a best efforts pattern for the future.
Their accuracy will obviously fall as they go further into the future.
Ultimately, the most that can be expressed in that the concept of DST
is likely to continue, not exact dates.

My approach would be that if a country/zone/ID has DST now and looks
likely to continue using it then the rules must end in a matched pair
of "max" transitions. I think such an approach would handle this case.

Personally, I would have only pre-generated the guesses for a shorter
timescale of say the next 10 years before going onto a generalised max
rule. But pre-generating to 2037 isn't "wrong", just a choice.

Finally, I note that applications are used to handling tz rules that
change in the future. My experience is that they are less well setup
to handle rules changes affecting the past, hence recent discussion.

Stephen


More information about the tz mailing list