New issue?: vocabulary maintenance

[Norman Gray]

Greetings, all.

Earlier today, after I'd posted the vocabularies issues list, Ed  
mailed me to say (reposting here with permission):

> I have comments on specific terms.  Some should be removed.  Some  
> should be replaced or altered.   Some should be broken up into  
> several terms, etc.  Where does that go?

This is an excellent question, to which I have no clear answer.

Up to now, we've been focused on getting the various vocabularies into  
SKOS one way or the other, but this has completely skipped the  
important discussion of how we actually maintain the vocabularies,  
both in the sense of discussing adjustments leading up to the  
standardisation, and establishing how they should be curated long-term.

There's a related issue, in the question of what status the  
vocabularies should have.  Chatting off-list last month, Rick and I  
seemed to be mutually and provisionally happy with the notion that  
what was being standardised here was the notion of vocabularies-as- 
SKOS, and that the vocabularies themselves were present as _examples_,  
and as a convenient repository for this initial set of important  
vocabularies.  However we never examined this closely, and it  
certainly hasn't been more widely discussed.  Now is the time.



It seems to me that the status of the vocabularies, and their future  
maintenance, are two dimensions which, though not orthogonal, are not  
degenerate.

[ For reference, the current version of the document refers to six  
vocabularies, namely the A&A journal keyword list, the IVOA AOIM list,  
the 1993 IAU thesaurus, an IVOA Thesaurus based on that, a UCD1+  
vocabulary (I have some quibbles in mind about whether that's a  
vocabulary of concepts, or a set of types), and a SKOS version of the  
list of constellations. ]

1. One extreme position is to say that the vocabularies distributed  
with the document are indeed just examples, which means they have no  
normative force, which means we don't have to worry too much about  
their coherence or accuracy, and we're not obliged to do anything with  
them post-standardisation.  In this view, the answer to Ed's question  
is: 'the IVOA doesn't care'.

I don't think this is a good idea.  Whether they're formally normative  
or not, if people are going to use these vocabularies (and we hope  
they/we want to, if the effort is to have any point), then they're  
going to use the versions which are included in the standard  
document.  We can't just wash our hands of this.

2. Another extreme position is to declare that all these vocabularies  
are normative, and they will be changed only by future revisions of  
this IVOA standard.  As a consequence, we will set up some bureaucracy  
to handle problems and reach consensus on modifications.  Here, Ed's  
answer is: 'fill in this form and the committee will get back to you'.

This isn't impossible, and the process which maintains the UCD word  
list could act as a model here.  How active is that process?  If it's  
not active, why not and does it matter?  Will the process in place for  
the single UCD list work smoothly when extended to six vocabularies?   
In any case, one goal of the vocabularies work (at least from my point  
of view) was to make it feasible to open up the process of vocabulary  
creation, and let users use whichever vocabulary best suited them or  
their community, and having a rigid list of standardised vocabularies  
might militate against this openness.

3. An intermediate position is perhaps to produce a set of six or  
fewer vocabularies which are reasonably good -- useful but not perfect  
-- but make a commitment _not_ to change them in any future version of  
this standard.  Instead, any group which wished to, could produce a  
separate standard (IVOA or other) which developed them, and maintained  
them by whatever mechanism it wished.  They needn't even produce a  
standard: an individual or institution could produce a variant or a  
completely new vocabulary, and perhaps publish it as an ADASS or A&AS  
paper.  As long as it had a distinct namespace, and the vocabulary  
file were available long term, like a online table in an A&AS article,  
this would be fine.

That answers half of Ed's question: 'in the future, ask someone  
else'.  It doesn't answer the more immediate question, of how do we  
produce these initial adequate versions?  Perhaps the simplest answer  
there is to identify some individual or team which will adopt each one  
of the vocabularies, and negotiate with them, on-list or off.   
Alasdair produced the A&A and AOIM vocabularies, Rick the IVOAT and  
constellation ones, Alasdair produced the IAU93 one from a variant of  
Rick's IVOAT script, and I confess to being unsure of the original  
authorship of the SKOSified UCD list, or how much of an exercise this  
was intended to be.

Perhaps this is the other half of the answer to Ed's question: can we  
have volunteers to adopt a vocabulary for a couple of months leading  
up to the standard?  If we can't find adopters, and the vocabulary is  
inadequate as it stands, then we can drop that vocabulary from the  
document.

Does that sound plausible?  Ed in particular: are any of those  
reasonable answers?

All the best,

Norman


-- 
Norman Gray  :  http://nxg.me.uk
eurovotech.org  :  University of Leicester