Referencing STC in VOTable: IDREF problem

[Mark Taylor]

On Tue, 16 Sep 2008, Francois Ochsenbein wrote:

> 1. http://www.ivoa.net/internal/IVOA/IvoaVOTable/note_stc_v0.8.pdf
> 	proposed note on "Referencing STC in VOTable"

Francois,

there is a problem with the scheme outlined in this Note.

Section 4 says that the ref attribute of the second required GROUP
element may reference an STC 'library' coordsys by using an
identifier along the lines "ivo://STClib/CoordSys".

This practice is exemplified in the Example in section 5.1:

    <GROUP ID="Coo1" utype="stc:AstroCoords"
    ref="ivo://STClib/CoordSys#UTC-ICRS-TOPO">

However, since the GROUP element's ref attribute is defined in the 
VOTable 1.2 schema with a type of "xs:IDREF" it is syntactically
restricted in what characters it can contain: in particular non-XML-name
characters such as "/" are illegal here.  Hence this example cannot
be valid against the VOTable 1.2 schema.
Even apart from the syntactical issue, I don't *think* that the ID/IDREF
mechanism can perform crossreferencing between different documents,
here the VOTable instance and the STC library, which is what appears 
to be intended here, though it may be that my understanding 
of that issue is wrong (can anyone else confirm/deny?)

The exact same problem afflicts the example in section 3.1 of the 
VOTable 1.2 document draft.

A rethink is required.  One possibility would be to keep the same 
general idea of referencing STC library documents, but to make it
informal rather than enforced by the ID/IDREF mechanism.
I recall however that Arnold was not keen on this approach when a 
similar solution was suggested to the problems arising from 
concatenating STC documents a year or two ago.

I have two more general comments:

    1. I would consider it good practice when writing a document such
       as this to validate any examples quoted against the schema prior
       to publication.  As well as the issue above, there are a number
       of other errors in the first example of the Note which validation,
       and in some cases even XML parsing, would have picked up
       (stray </FIELD> close tag, missing IDs for the 'second referencing'
       refs).  I haven't checked the other examples.

    2. The approach outlined by this Note is quite complicated, though
       perhaps that is unavoidable given the job it is trying to do.
       Even assuming the problem I have identified above will be fixed,
       I am not therefore very hopeful about seeing applications capable
       of making sense of VOTables containing coordinate data in this form.
       I would of course be quite happy to be proved wrong about that.

Mark

-- 
Mark Taylor   Astronomical Programmer   Physics, Bristol University, UK
m.b.taylor at bris.ac.uk +44-117-928-8776 http://www.star.bris.ac.uk/~mbt/