Comments on SSAP V0.95, UCDs

[Doug Tody]

The data models also have UTYPEs (you will see these in serialized data
that comes back for example), but an interface specification or protocol
supercedes these and specifies all aspects of the interface directly,
for just the sort of reason you refer to.  For a given version of the
protocol anything like this is fixed, and fully defined (the VOX UCDs
in SIA 1.0 are an example of this).  The only caveat is that one could
in principle "pass through" component data models in a query response,
but we don't currently do that; the interface is fully defined, in part
because the data models are evolving separately.

Note, cone search is different.  Since there is no data model for general
tables (at least not currently), all you see are the UCDs.  The closest
analogue to UTYPE for a general table is probably the field name assigned
by the creator, which probably does uniquely identify each table field.

	- Doug


On Tue, 27 Jun 2006, Randall Thompson wrote:

> Oh sorry,  I thought the UTYPEs were dependent on the underlying
> data models. So if they can be defined and required for all VO protocols,
> there is no problem. Thanks.
>
> Randy
>
> Doug Tody wrote:
>
> >Hi Randy -
> >
> >The purpose of UTYPE is to identify interface elements; they are intended
> >to be used for this purpose in SSA, and the SSA protocol will specify all
> >the UTYPEs independently of the underlying data model.  UTYPE replaces
> >the "VOX" UCDs used in SIAP, and the next SIAP will use UTYPE as well.
> >UCDs can also be specified, but different interface elements may have the
> >same UCD so these cannot be used to identify interface elements.
> >
> >	- Doug
> >
> >
> >On Tue, 27 Jun 2006, Randall Thompson wrote:
> >
> >
> >
> >>Hi Doug,
> >>    Thanks for the reply. I hope to discuss
> >>the issues further with Bob and the MAST staff
> >>when everyone is back from travel.
> >>    One issue I forgot to ask about was the status
> >>of UCDs in the SSAP. Currently our service-handling
> >>software relies on UCDs to identify the main RA and Dec
> >>values, data links, etc. and works with both cone searches
> >>and SIAP requests. I was hoping the same UCDs would
> >>be required for the SSAP so we could use the same code for
> >>all three protocols. Since UTYPES can vary with the
> >>data models, they would not be as useful for this purpose.
> >>Is it possible that we can continue the requirement of
> >>specifying UCDs for at least the more important SSAP
> >>parameters?
> >>
> >>Randy
> >>
> >>
> >>
>