FW: dateTime Schema counter proposal

Michael S. Brothers [mailto:Michael.S.Brothers@EMCIns.Com] wrote
>>Subject: Re: dateTime Schema counter proposal
>>
>>This is a great suggestion. May I add that while this spec is defined, 
>>we go ahead and take care of the Y10K/Y100K problem at the same time?

The most obvious way to do that to me is to allow a secondary lexical
representations for date and dateTime (ala timeDuration) that matches the
real production and is interpreted like Microsoft's DATE.

From MS Documentation:
The DATE type is implemented as a floating-point value, measuring days from
midnight, 30 December 1899. So, midnight, 31 December 1899 is represented by
1.0. Similarly, 6 AM, 1 January 1900 is represented by 2.25, and midnight,
29 December 1899 is – 1.0. However, 6 AM, 29 December 1899 is – 1.25.

This actually takes care of all arbitrary time boundaries (most
implementations would have a year 5E307 problem if based on a IEEE double)
though you really start losing precision a couple of trillion years from
now.

I had thought about it adding it originally, but I wanted to avoid anything
that would distract from the rest of the proposal.