I'm not sure whether we really need "Year and ordinal day", but if we are
to discuss this question, there is something important to keep in mind:
> January 3 of 2003 can be represented as 2003-01-03.
Yes, as long as the day number is below 60 and / or it is known whether
the year is a leap year or not, which is not always known. Let's consider the
The representation of y30000-070 as year-month-day depends on whether year
30 thousand will be a leap year or not, which is not determined yet,
because of the limitations of the gregorian calendar, which (as it is
defined today) is grounded on a coarse approximation of the number of days
needed for the earth to rotate around the sun. This approximation will
need revision (within a couple of thousand years or so). My message is the
following: if we choose to disregard such remote years, this should be
expressed in the specification. If we choose not to disregard such remote
years, then we should note in the specification how we solve this problem.
The easiest (at least in a short term) would be to remove "Year and
ordinal day" from the specification now and maybe try to solve this
question in a later version of the specification.