From: Saašha Metsärantala
> - For clarity, I would suggest to split #004 into two parts: #004 for
> intervals containing "date / date" and #005 for intervals containing
> "date / period".
> - In the "examples" column, all "date / date" intervals are symmetrical,
> that is with year-year, month-month or day-day precision. In the
> column, other constructs are described, where the year (and sometimes
> even the month) is removed. I consider that symmetry (in this context)
> has the advantage of clarity and that other constructs (if we want them)
> should be described in detail when it comes to their semantics.
Yes we had this discussion a few weeks ago and concluded that there was no need to support the asymetrical representation and that is was ambiguous anyway. I've changed this.
> - According to the "rules" column, it is not clear whether we should
> which I consider could be shortened (and canonicalized) as
> where M's and D's are assumed to be zero when absent from such a
I don't know what these sort of normalization rules should be for duration, and I welcome comments.