Re: [Tzdist] How should tzdist support rscale?

[Jonathan Lennox <[e-mail]>]

On Tuesday, January 6 2015, "Cyrus Daboo" wrote to "Jonathan Lennox, [e-mail], [e-mail]" saying:

> Hi Jonathan,
> 
> --On January 5, 2015 at 5:01:02 PM -0500 Jonathan Lennox 
> <[e-mail]> wrote:
> 
> > A question recently occured to me about the interaction of rscale and
> > tzdist.
> >
> > Should it be possible for a VTIMEZONE -- either inline in an iCalendar
> > document, or distributed by tzdist -- to use an RSCALE?
> 
> If a time zone were indeed defined in terms of a non-Gregorian recurrence 
> then it would seem reasonable to support that.
> 
> > There are some natural use cases for this -- Iran uses the Persian
> > calendar to set its DST observances, and Israel used the Hebrew calendar
> > from 2005-2012.  Right now the IANA tzdata just expands non-Gregorian
> > rules to a list of dates, but in principle, RSCALE would be a much more
> > natural (and compact) way of representing rules like these.
> >
> > However, tzdist has no way to negotiate the use of iCalendar extensions.
> > Should one be defined?
> 
> If tz data publishers actually start wantint to produce data using 
> non-Gregorian rules, then we can easily define an extension to the 
> capabilities response to indicate that is an option.
> 
> > For that matter, the rscale draft doesn't talk about its use in VTIMEZONE
> > at all (indeed, it asserts that VTIMEZONEs are always specified in
> > Gregorian time).  Should it?
> 
> Actually the rscale document only mentions time zones once in Section 1 - 
> and there it states that time zone rules (i.e., what comes from a 
> publisher) is specified in terms of the Gregorian calendar - that does not 
> imply VTIMEZONE itself is limited to Gregorian rules - just that no time 
> zone data uses non-Gregorian. If you think that text needs re-wording to 
> clarify that, I can certainly change it.

Yes, I think that the rscale draft should be clarified to note that RSCALE
can appear in an RRULE in any iCalendar context where RRULE is allowed, not
just VEVENT, and in particular it can appear in VTIMEZONE.

> At this point, I think the use of non-Gregorian rules in VTIMEZONE should 
> not be ruled out, but tzdist does not need to support it until we have 
> publishers actually wanting to use it. Of course at that point things will 
> get tricky for the publisher as they will want to keep the Gregorian rules 
> around for backwards compatibility.

Yes -- for tzdist I'm mostly thinking about this as an exercise to make sure
we have backward-compatible future extensibility defined correctly.  If we
understand how we would do this if it's ever needed, I think we're in pretty
good shape.

That said, I think tzdist in the meantime should have some statement saying
that RSCALE MUST NOT be used unless support for it is negotated through some
as-yet undefined mechanism.

-- 
Jonathan Lennox
[e-mail]