UCD comments 3: the standard vocabulary for the universe
Rick Hessman said:
: : Sebastien Derriere said:
Actually, it was me who said this, Sebastien's message quoted me.
At the time I wrote this, I actually agreed with Rick; I was trying to find a compromise approach since it seemed that Sebastien (and I think Patricio and other originators of the UCDs) do want to make such a distinction. Someone is going to have to list object-classification concepts. I would prefer it was the UCD board, but if not, it'll have to be the DM group or the vo-semantics folks.
However, after the discussion I take a more nuanced (i.e. confused) position.
I do now think there is some real justification for a separate list using the
ucd = src.class value = galaxy.spiral
approach: the other approach with
ucd = src.class.galaxy.spiral
might be ok, but very quickly we want
ucd = src.class.galaxy.spiral.SB(r)cd... ucd = src.class.star.binary.O7e+PNe/LMXB ucd = src.class.comet.sungrazer ucd = src.class.body.feature.rupes
In other words, 'galaxy' is not just a concept, it's a classification,
and the coarsest bin in a complicated tree of classifications.
(Sebastien made a distinction between classification and morphological
type which I don't think is a real one in practice; morph type, variable
star class, spectral type, etc. are all part of the classification of
what this object is).
I don't think the classification's really quite the same as
distinguishing between radial velocity and some other kind of velocity.
And as my examples above show, once you get down to the detailed
classification the UCD syntax doesn't actually make a good match. Rick
will probably argue that it still makes sense to have UCDs which go down
as far as the less detailed level (star.binary etc) and I could buy that.
Like Rick, I focus on the data-header applications more than the Vizier-catalog applications, and I take it as a clue that in FITS headers, "spiral galaxy" is always a keyword value, not a keyword name. This suggested to me that in fact UCD may not be the right controlled vocab for this set of concepts.
But then Sebastien reminded us of the 'concept/property' way of looking at UCDs, where the question becomes: are there specific properties we need to describe? To be more specific:
radial.velocity;src.class.galaxy
would be the radial velocity of a galaxy. Does having this add anything to just saying
radial.velocity;src
This, to my mind, is the correct role of object types in UCD and argues for a tree of this kind. Right now, these kind of properties go in the phys or src list without the 'object.<particularkind>.' prefix.