Expressing position in RDF

[Rob Seaman]

Matthew Graham wrote:

> Although this works, it is very verbose for such a simple (!)  
> statement.

Actually, this is rather terse for your typical STC use case (which  
this is).

> To sell this sort of stuff to end users, we really need a practical  
> and succinct set of object properties

Practical and succinct are often at odds.  This is a corollary of  
Gödel's theorem - by analogy, but likely also with mathematical rigor.

> that would allow me just to do:
>
> :MyObject a:rightAscension 134.556^^xsd:float
>
> After all anything observable in the sky has a RA.

There are several ways in which that assertion fails.  RA is always a  
function of time - sometimes slowly varying, sometimes rapidly.  The  
thing that "has a" RA is the combination of the observer and the  
object.  Things that have an RA are also required to jointly have  
other properties (Dec, type of coord system, an equinox for some of  
these systems).

Further, not all "things" are negligible in extent, so an area  
(footprint if you will) may need to be conveyed.  And the object  
doesn't really "have" the RA in any event, but rather this is simply a  
happenstance of the point of view (unlike, say, the temperature of a  
star).  Which is all to say that this is a property of a detection  
(i.e., in an instrumental catalog) rather than of an object (later  
deduced from instrumental observations).

Consider, what is observable must be a physical process of some sort,  
not a Platonic ideal of an astronomical object.  For example, if on  
the surface of a planet or star, what is observable may periodically  
vanish from view.  The observable may also involve interactions of  
multiple objects.

In VOEvent-speak, <What> describes assertions about the object(s)  
itself, but targeting coordinates in <WhereWhen> are assertions about  
either a past or future, real or abstract, observation "on the" (set  
of) objects.

Which is to say that Ed and I were trying to do you a favor by  
simplifying it all to "object has a WCS" and "WCS has a RA" :-)

Another way to look at all this is simply E-R DB-style normalization.   
You don't want to flatten the data structures too much.  Ultimately  
only so much intrinsic complexity can be hidden from end users.

Rob