> we need only to represent days, months, and years.
As of XML Schema 1.1, "day" as a duration is nothing else than a multiple 
of seconds. Any duration written in days may be rewritten to seconds and 
the other way round: just divide or multiply by 86400. That's how 
datatypes in XML Schema 1.1 works. I think that the concept of precision 
may be useful for us and I would like to make use of it, but I would 
prefer to formulate it another way and keep the concept of day duration as 
of XML Schema 1.1 so that data exchange will be made easier.

> But is "2 years, 5 days" really meaningful?
According to XML Schema 1.1, this would mean a duration of 2x12 = 24 
months (on the month dimension) plus 5x24x60x60 = 432000 seconds (on the 
second dimension). I think it could make some sense in


which can not be easily rewritten as something else.

> So, the normalization rules would be different for a standalone duration
> than for a contextual duration.
Yes, of course!

> does anyone see a requirement for standalone duration?
This question must be asked, of course! I was just assuming that we were 
discussing standalone durations. Contextual duration can always be reduced 
to a number of seconds (and its multiples such as days, etc.) if the 
start-point is known with a granularity no lower than that of the 

I would like to ask whether we need CONTEXTUAL duration AS OF #005? Is 
there a point with a start-duration interval (as of #005) if it always can 
be rewritten as a start-end interval as of #004? Of course, contextual 
duration may be useful in expressions such as:


but there is no example for that in the EDTF specification. Therefore, I 
wonder whether we could move CONTEXTUAL durations to level one where they 
could be useful (for example when the start-point is combined with "u"). I 
do not find contextual durations useful at level zero as they are defined 
in #001.