RE: Questions regarding recurrence rules

["Laird Nelson" <lairdnelson@xxxxxxxxxxx>]

> -----Original Message-----
> From: owner-ietf-calendar@xxxxxxxxxxxx 
> [mailto:owner-ietf-calendar@xxxxxxxxxxxx] On Behalf Of 
> Reinhold Kainhofer
> 2) All VTIMEZONEs (at least those used by Exchange, Apple's 
> iCal, and by libical and so by evolution, korganizer etc.) 
> use RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU

(Mr. Kainhofer is implying that they got it wrong; something with which I
agree.  Mr. Roberts points out in another email that really, yes, that RRULE
should read RRULE:FREQ=MONTHLY;BYMONTH=10;BYDAY=-1SU, with which I also
agree.)

However, consider this timezone definition excerpt from RFC 2445 (line 3719
in the .txt version):

     BEGIN:STANDARD
     DTSTART:19671029T020000
     RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
                ^^^^^^
The RRULE here is identical to the "wrong" way of doing it that Mr.
Kainhofer describes above, and probably explains why Microsoft and Apple and
the libical folks all got it "wrong" in the same manner (i.e. because the
"wrong" way is the one that is given as an example throughout RFC 2445).
Apparently this is the way that RFC 2445 encourages you to write RRULEs
featuring integer-preceded BYDAY modifiers in conjunction with other
modifiers.

If this is correct and can be generalized, then it implies that BYDAY
modifiers apply to the current set of evaluated occurrences, not to the
FREQ.  This contradicts what Mr. Roberts was saying earlier.  The upshot is
that I'm very confused.  Perhaps my recurrence rule expansion library will
need to be written to operate in one of two different modes?

I am assuming from the relative silence on all of these discussions that Mr.
Dawson and Mr. Stenerson are no longer on this list and cannot chime in to
offer the framers' intent.

Best,
Laird