New issue?: vocabulary maintenance [vocabset-5]

[Frederic V. Hessman]

On 6 Feb 2008, at 6:02 pm, Norman Gray wrote:

>> UCD (an informal suggestion of how to standarize UCD keyword  
>> publishing),
>
> I've cast doubts before about the appropriateness of including a UCD  
> vocabulary in this exercise, but I'd like to recant them.  The  
> vocabularies other than UCD are describing things that astronomers  
> study, whereas UCD is describing data types, which is a very  
> different thing.  But even though it's unlikely there will be  
> mappings from a SKOSified UCD to the other vocabularies, the UCDs  
> are still a vocabulary, so no more quibbles from me on _those_  
> grounds (but see below).

There's a very simple reason why there needs to be a SKOS-ified  
version of UCD:  we desperately need to be able to link vocabulary  
words to UCD tokens.  If we can't do this, then we'll need to replace  
all that UCD is and put an equivalent in a different vocabulary (e.g.  
IVOAT), which sounds like a pretty stupid thing to do.

>> and IAU-93 (on historical principle).   The IVOA-T is definitely  
>> not ready to be called "core" and we purposefully left it out of  
>> the list.
>
> ...which is why you titled the relevant section `Towards an IVOA  
> Thesaurus' -- I see.
>
> There's a bigger range of expectations here than I thought.  I had  
> thought that the idea of what we have come to call the IVOA-T was  
> that it would be essentially the IAU-93 thesaurus with some of the  
> more egregious problems straightforwardly updated.  That is, it  
> would be quite a small project to produce the first public version  
> of the IVOA-T, viewed as an updated version of IAU-93.
>
> Some here, however, clearly see it or have come to see it as  
> involving much more (in some cases very much more) substantial  
> updating and reworking.
>
> Would it be reasonable to produce such a basic update in the couple  
> of months we're aiming for?

If more people than Ed and I worked on it......

> The two options seem to be:
>
> 1. Publish just the IAU-93, but with a prominent notice saying we  
> know it's out of date, and this is just an exercise (the  
> disadvantage is that this starts to look redundant, or alternatively  
> that the fact that `everyone knows it's out of date' might partly  
> undermine the standard).

Even if it is much less useful than it should be, it's not "out-of- 
date", if only because it still is the international standard, much  
more than AOIM.

> 2. Publish the IAU-93 plus a minimally updated IVOA-T, with a note  
> saying that this is not intended to be complete, but that it  
> nonetheless has the prime virtue of existence, and can act as the  
> basis for a fuller, separate, improvement effort.  This route has  
> some precedent in the way that UCDs first appeared.
>
> My vote would be for (2); Rick, you'd go for (1), yes?

The present IVOA-T IS a minimally updated IAU-93, but "minimally  
updated" doesn't simply mean we've added a few missing words.  There  
is some concern that a discussion of the merits of publishing an IVOA- 
T will distract from the expected easy acceptance of the basic  
vocabulary proposal.   On the other hand, if we tell our colleagues  
"be fruitful and let vocabularies multiply in the VO" but then say  
that the only ones available at first, A&A, AOIM, and IAU-93, aren't  
really enough to cover the basics, then they'll rightly say "why  
bother" or simply use IAU-93 after all (like Rob).

>> Keep the initial list short.  Our goal isn't to anticipate all  
>> future participants, rather to provide a mechanism that will permit  
>> new (or quite mature) vocabularies to be added when needed.
>
> I suggest that we include in the Vocabulary standard document the  
> three vocabularies A&A, AOIM, and either IAU-93 or IAU-93/IVOA-T,  
> depending on how we resolve this.

This is a good point:  the vocabularies we discuss at more length in  
the standards document don't have to be the same ones which the IVOA  
publishes.  Still, if we can produce several useful vocabularies (each  
having it's own particular use) and we're not stepping on anybodies  
semantic toes, why can't we simply produce a nice batch of  
vocabularies to start out with?
>
> I suggest omitting the SKOSified UCDs on the grounds that their  
> subject is conceptually separate, and in any case they might be more  
> rationally published via the UCD maintenance process.

No, again because we need to be able to like vocabulary tokens with  
their UCD equivalents.

> I don't know what to do with the constellation vocabulary.  On the  
> one hand, it's simple; on the other, 4 is 33% more than 3....

Exactly what we now have:  it is a usable if limited vocabulary  
derived for didactic purposes, being MUCH simpler than all the rest  
we're taking about.

>> Is there a practical limit to the number of vocabularies we will be  
>> able to support?  There may ultimately be dozens or hundreds (or  
>> more) distinct dialects.
>
> The practical limits will I think be twofold: can UIs cope with this  
> range in any helpful fashion?; and can the community collectively  
> manage to curate this many vocabularies and their mappings without  
> them all going stale?  I don't believe there are important technical  
> limits.  I can imagine a dozen vocabularies, some of them rather  
> small.

Let Darwin solve this problem.

>> This raises, yet again, the point of finding a permanent place  
>> within www.ivoa.net to put these things.....
>
> I'm glad you raised that. I don't think there's a deep problem here,  
> but I'm not sure where it should be discussed.  I imagine I just  
> mail Bruno (the document coordinator).

No, because this should be the long-term place for all IVOA-supported/ 
mirrored vocabularies.  This demands a long-term commitment to a  
simple URI not unlike http://www.ivoa.net/xml or http://www.ivoa.net/Documents 
.

Rick