[Cube/vo-dml] ivoa datatypes

[Gerard Lemson]

Hi Rob 
> On Apr 28, 2014, at 12:19 PM, CresitelloDittmar, Mark
> <mdittmar at cfa.harvard.edu> wrote:
> 
>>
>> On Fri, Apr 25, 2014 at 4:55 AM, Gerard Lemson
>> <lemson at mpa-garching.mpg.de> wrote:
>>
>> > Request:
>> >
>> >   During this effort, there are a couple types that I would like to
>> add to this suite.
>> >
>> >   1) TimeValue: extends AtomicValue
>> >
>> >         + value: datetime     # ivoa:datetime primitive type
>> >         + unit: Unit              # ivoa:unit primitive type.
>> >
>>
>> Arnold has made some comments about different types of Time value object
>> depending on whether this is absolute or a delta (time since T0).
>> Part if my trouble with debating this is not knowing the scope of
>> 'datetime'.  Is there an actual specification of what this covers?
>> If this is just a time instance type (suitable for DATE keyword), then
>> it may not be what is desired under Quantity.
> 
> Perhaps start with use cases.  "Time" is arguably the broadest topic in
> existence.  Before a specification, one needs engineering requirements. 
> Before requirements, a concept of operations.  What is it you (the
> collective "you") are trying to represent and for what purpose?
> 
The direct use case for which we added the various types in the
ivoa:quantity package in the 'ivoa' model (in
https://volute.googlecode.com/svn/trunk/projects/dm/vo-dml/models/ivoa/IVOA.
vo-dml.xml) was so that in the vo-dml/utype specification, we have a
predefined set of special VO-DML DataTypes that can be directly represented
by a VOTable FIELD or PARAM. 
These types have a 'value' attribute of a certain primitive datatype and
possibly a ucd and/or unit. The assignment/mapping of a VO-DML Attribute
with such a datatype to a FIELD or PARAM would imply that its 'value' is
mapped to the TDs or @value attribute respectively, and its 'ucd' and 'unit'
to the corresponding attributes in the VOTable metadata element.
So far in the ivoa model we defined special quantity types for the primitive
types 'integer', 'real', 'boolean' and 'string', but not yet for 'datetime'.
I think Mark's suggestion is to add such a type. 

One might wish to discuss the issue that VOTable does not have its own
datatype for representing dates and/or times, but I think that is not a real
problem as the utype spec does not mandate that attributes are mapped to
FIELDs/PARAMs of compatible types. 

Main issue I think is whether a single ivoa:quantity.DateTime (for example),
representing a (datetime,ucd,unit) tuple, can represent all(?) possible ways
in which one might wish to express date/time quantities. 
I very much hope this can be resolved by time domain experts in the ivoa.

Cheers
Gerard
> Rob
> 
>