[TAP] Summary: data type for column metadata

[Rob Seaman]

Mark Taylor wrote:

> - I'd say TIMESTAMP as a char array using ISO-8601 format makes  
> sense.  But I may be prepared to retract that if Rob or Arnold bites  
> my head off.

I've got nothing against a restricted subset of ISO-8601.

My earlier point was that "timestamp" is not an overarching concept  
that applies to every possible timekeeping situation.  Sometimes you  
do indeed want an epoch, the concept which appears to most closely  
correspond to the notion of a timestamp that has been tossed around in  
this forum.

Sometimes you want a calendar date.  The lack of a time does not in  
this case represent a 24 hour imprecision, rather today is 16 April  
2009 all day long.  The handling issues (e.g., localization and  
logistics) may be different than with timestamps.

And sometimes you want a clock time, e.g., an expression of Earth  
orientation, without a date.

The latter two cases can also be handled by ISO-8601 or sexagesimal or  
floating point or even integers.

Rob