VOTable 1.5 Working Draft (2023-09-13)

[Markus Demleitner]

Hi Pierre,

On Mon, Sep 25, 2023 at 06:02:52PM +0200, Pierre Fernique wrote:
> which referred to the IVOA note "2009 - Referencing STC in VOTable v1.1" => https://www.ivoa.net/documents/Notes/VOTableSTC/NOTE-VOTableSTC-1.1-20090612.html,
> then v2.0 (2010) (see example C) with the difference that the COOSYS had
> been replaced by a GROUP and different utypes have been proposed. As a
> reminder, this method was subsequently deprecated because no VOTable
> provider had implemented it - even several years after its adoption. Version

Well, there were two, but yes, adoption certainly is a concern, which
is why the present proposal is accompanied with an astropy
implementation <https://github.com/astropy/astropy/pull/14992>
[comments and help welcome!] that will let do people automatic epoch
propagation, a feature I'd hope is reason enough for implementors to
take up the new mechanism.

> This proposal in 1.5 is more flexible than version 1.2 (which completely
> eliminated the use of COOSYS). The fact remains that having 2 association
> methods still poses consistency questions? Which is required, which is
> optional? What should be done when the 2 methods are simultaneously present,
> but not consistent? Should we expect the original referencing to disappear
> (FIELD ref -> COOSYS ID)? What are the consequences? what are the benefits?
> Will this new referencing method be extended to TIMESYS? ...

You may remember that the original TIMESYS draft also used this
"forward" refrencing, so I'd say, yes, *if* we keep COOSYS and
TIMESYS around once MIVOT is there, we should transition both to
mandatory forward references in the end, just like MIVOT itself does
things.

For version 1.5 and the next few VOTable version, I think it would be
rather hard to exactly specify what is mandatory under which
circumstances -- not even the backward referencing with @ref is
mandatory at this point, and we probably wouldn't want to make all
usage of COOSYS without forward references illegal in one go anyway.
So, as much as I'd like to require forward references mandatory, I
don't think that'll be prudent any time soon.

For now, I consider the COOSYS references a bridge to later full
MIVOT+Coords; implementors have something "simple" to start adopting
MIVOT-style annotations, so once services start publishing full MIVOT
annotations they will only have to adapt to a (not terribly
different) syntax but have a solid core functionality in place
otherwise.

> The discussion on this evolution of VOTable 1.4 -> 1.5 - which may seem
> minor, but can have major consequences - can only be found on github
> (started at the end of June 2023 =>
> https://github.com/ivoa-std/VOTable/pull/40), and relatively few people have
> reacted since then. Would it be possible to use the Apps IVOA mailing list
> to reach a wider audience.

Certainly.  Thanks for bringing this up in an open forum.

          -- Markus