Re: [calsify] RFC 5545/JSCalendar questions from ECMAScript TC39 "Temporal" working group

[Justin Grant <[e-mail]>]

BTW, "But if E falls inside a DST skipped hour" should have read "But if
END falls inside a DST skipped hour"

On Tue, Sep 29, 2020 at 4:43 PM Justin Grant <
[e-mail]> wrote:

> Hi Neil - I'm working on a sample implementation of the algorithms we've
> been discussing for addition, subtraction, and calculating the difference
> between date/time values.
>
> I'm trying to figure out if there's an algorithm that can make all
> operations reversible, meaning that the last two comparisons below should
> be true:
> * let START = some date/time value with a time zone
> * let D = some duration
> * let END = START + D
> * let DIFF = END - START
> * let UNDO = END - DIFF
> * assert DIFF == D
> * assert UNDO == START
>
> But if E falls inside a DST skipped hour, then I'm not sure that both the
> comparisons can be true. Example:
>
> * START = 2020-03-07T02:30-08:00[America/Los_Angeles]
> * D = P1D
> * END = START + D == 2020-03-07T03:30-07:00[America/Los_Angeles]
> * DIFF = END - START == P1D (OK)
> * UNDO = END - DIFF == 2020-03-07T03:30-08:00[America/Los_Angeles]
> * UNDO != START (3:30 vs. 2:30)
>
> I think this result is probably OK (because DST is weird) but I'm
> wondering if the difference calculation in this case should return P1D
> (which makes `end - start == P1D`) or P1DT1H (which would make `undo ==
> start`). Or is there some possible algorithm which could make both
> comparisons true?  If not, then would it be better for the first comparison
> to work and the second to fail, or is it better for the second to work and
> the first to fail?
>
> Cheers,
> Justin
>
>
> On Sun, Aug 16, 2020 at 4:39 PM Neil Jenkins <[e-mail]>
> wrote:
>
>> On Sun, 16 Aug 2020, at 05:16, Justin Grant wrote:
>>
>> Hi Neil - reviewing this now. Had a quick question:
>>
>> > Let T = Y - X in absolute time => PT23H45M
>>
>> Why is X used here instead of I?  Here's what I would have expected:
>>
>> Let T = Y - *I* in absolute time  => PT23H45M
>>
>>
>> Correct. This was an error, it should be *I* as you say.
>>
>> Neil.
>>
>