[kepler-dev] Float, Byte and Short types in Ptolemy

[Tobin Fricke]

On Tue, 27 Jul 2004, Ilkay Altintas wrote:

> Also, having an XSD-like Date type in the type system would be very
> useful especially in the context of Kepler web services and dataset
> discovery part.

Having a 'date' type is fairly important for geophysical applications (to
say the least!).  There is some subtlety in defining a date type -- the
usual approach is to describe a point in time as the number of seconds
(expressed in a floating point format) since a particular 'epoch'
(1970-01-01 00:00 in the case of Unix).  Issues like leap seconds, leap
days, timezones, and calendar oddities make the definition of time a
little more complicated.  Having a standardized date format would be quite
useful -- we should ensure that it's a format that's useful for both
engineering and science purposes.

I think the seconds-since-epoch-with-respect-to-UTC format is good since
it avoids dealing with most calendar oddities (although not leap seconds).
I don't know anything about XSD, so maybe they've dealt with this already.
The Java standard library defines a few date/time/calendar routines, but
their specification seems a bit ambiguous on issues such as timezones,
etc, and might even defer to the host operating system.

Tobin