Re: [calsify] SKIP was Re: AD review of draft-ietf-calext-rscale-03

[Gregory Yakushev <[e-mail]>]

On Thu Feb 05 2015 at 4:49:39 PM Barry Leiba <[e-mail]>
wrote:

> > As mentioned in another thread, I'd prefer to have SKIP=BACKWARD as
> default
> > to reduce the total amount of text in RRULEs (that is, sum of all RRULEs
> > stored in the world). Length of RRULEs matter from both performance and
> > readability point of view. But SKIP=OMIT is an acceptable compromise if
> > consensus points in this direction.
>
> And yet I'm still concerned that trying to define SKIP rules outside
> the context of implementation of a specific calendar is complex and
> error prone.  It remains my contention that the best thing to do
> (which also keeps the rules short) is to say that how RRULEs that
> contain RSCALE work without SKIP is dependent upon the calendar
> specified in RRULE, and that when you implement a particular calendar,
> you have to know how that calendar works and support it properly.
> Then we reserve SKIP for exception situations.
>
> I do NOT think that we need to accompany this with any list of
> calendars.  Whoever implements a particular calendar needs to know how
> it works in order to generate the right SKIP values anyway.  So why
> not just say that, and forget about having to include the SKIPs?
>

Because there should be an agreement between all implementation on what
SKIP default is applied to a given Calendar. Say, we agree here that SKIP
should default to BACKWARD for all calendars except HEBREW, for which it
should default to FORWARD. This information needs to be written somewhere
and set in stone, so that all implementation apply correct default when
expanding RRULEs.

Otherwise an implementor without deep knowledge of a particular calendar
(its not required if he bases his implementation on ICU for example), can
apply a different default. I don't see how can we do without a registry for
variable defaults, if we choose this option.


> Barry
>

Grisha