Voc in the VO 2 / remarks

[François Bonnarel]

Hi Rick, Markus, all

Le 03/06/2020 à 11:23, Markus Demleitner a écrit :
> Dear Frederic,
>
> On Wed, Jun 03, 2020 at 10:31:11AM +0200, Frederic V. Hessman wrote:
>>> On 3 Jun 2020, at 09:27, Markus Demleitner <msdemlei at ari.uni-heidelberg.de> wrote:
>>>>        2 ) for the  same reason (incomplete discussion) "Preliminary" terms
>>>> should not be set in the "official" vocabulary list even with the
>>>> "preliminary" tag. Examples using those terms should be provided in
>>>> prototype services or realistic example files.
>>> [...]
>>> Plausible examples IMHO aren't enough -- if there's no immediate need
>>> for a term, we should wait until that immediate need comes up, at
>>> which point we probably better understand the problem we're trying to
>>> solve.
>> I don't agree.  François is thinking of Voc primarily as a system
>> of ontologies - of data models - which of course need to be tightly
>> constrained in order to be useful.  I agree with Markus that we
>> need something simple and quickly useful.  For that reason, I think
>> our goal has to be to support vocabularies as simple term lists
>> with or without much ontological frill.  The usefulness of
>> hierarchical (avoiding the term "ontogolical") information within a
>> vocabulary depends on the use case.  A term list is most useful
>> when the end-user easily finds the terms she needs.  If a useful
> Well, I'd say "it depends".  See
> http://ivoa.net/documents/Vocabularies/20200326/WD-Vocabularies-2.0-20200326.html#tth_sEc2.1
> for the use cases that motivate much of this.  For one, some of them
> really need trees (as rather strict is-a hierarchies) -- and in many
> cases (think time scales or reference positions) not finding what
> you're looking for is a rather clear case compared to "soft" cases
> like the UAT.  When people are sure there's a hole in this way, I
> strongly suspect it's more likely they can be bothered to put in a
> VEP (or at least a mail so others make a VEP from it.

I was not exactly thinking to a system of ontologies, but hierarchies of 
terms do make sense for the two standardized list of terms I know well : 
dataproduct_type (there are obviously subtypes we may have to 
standardize) and DataLink "semantics" (where we already have several and 
wil have more).

The VEP as they are proposed at the moment are too  "strict".

When a vocabulary discussion starts on a use case you may not know if 
this will be solved by a term of a set of related terms.

You start with a single term and realize taht you need two parallel ones 
with subterms.

Cheers

François


>
>> term isn't present, the motivation to search for another vocabulary
>> is modest (unless we provide simple tools that do this), the
>> motivation to put up ones' own VO-compatible vocabulary as a fix is
>> highly limited and the expectation that some external gremium will
>> quickly add a term when asked is zero. This is why I wasn't so
> And do you think these expectations cannot be changed?  Not even,
> say, with some sufficiently prominent "Haven't found what you need?"
> button in a corner of the browser?
>
> I still have to confess that I've not really considered too much
> the scenario that people unfamilar with the IVOA might find
> themselves having to author VEPs.  I'd still rather think hard about
> ways to make it easier for them to participate than to try to
> anticipate their needs (and quarrel a lot about the anticipations and
> fail in the end after all).
>
>> happy when UAT turned out to be highly stripped down from the
>> original IAU terms list.  If terms are rarely used - no, my VO app
>> doesn't need the term "armillarly sphere" - does your computer
>> really care?
> The computer probably won't, but annotators will if there is too much
> ambiguity ("do I use #progenitor or #upstream-in-provenance?"), and,
> worse, several of these vocabularies will be reflected in code (e.g.,
> transforming reference positions), and there, it's critical that it's
> well understood what a meaning really is, and that these meanings are
> as well tried-and-tested as we can make them.
>
> Perhaps our small, very formal vocabularies really *are* something
> very different from big, soft ("SKOS") vocabularies and the two
> things shouldn't be regulated by a single document, or at least in
> two very different ways.
>
> But I'm still very convinced that the IVOA's general stance "show
> implementations if you want something to be standardised" is a really
> great idea.  I'd still hope that we can take that to vocabularies as
> well.
>
>          -- Markus
>