Laurie's UCD paper

[Jonathan McDowell]

Dear IVOA theory,
 after recent work on UCDs I've gone back to look at Laurie's
paper on UCDs for simulations. 
 In general I like the sim.* tree, but I don't like the
sim.param.* and phys.param.* trees. It seems to me that the
things listed in those trees are quantities that could simply
be in the phys.* trees, and might appear in other contexts than
simulations. I suggest instead sim.* as "UCD secondary words", and
so UCDs like
     phys.cosmology.hubble
     time.step;sim.param
     phys.size;sim.extent;sim.param
and so on. "What is this? First of all, it's a physical size. What
kind of physical size? It's the extent of a simulation, and it's one
of the parameters to the simulation".
 Similarly, rather than sim.obj.star I'd say "src.class.star;sim.obj"
("It's a star; oh by the way it's a simulated star") and
"pos.frame.cartesian" instead of "sim.coordsys.cart" (nothing particularly
"sim" about it, except that it's what STC calls a 'relocatable' frame - 
if you insist, "pos.frame.cartesian;pos.sim.relocatable")

In contrast, the sim.alg and sim.resource trees don't correspond to 
to astronomical things, but sim.resource UCDs could equally well correspond
to observation reduction pipeline things. An earlier attempt to propose
a software tree got shot down, but I think it has to come back at some point,
e.g.  soft.processors or meta.compute.processors or something. 

For the phys.param, how about
    phys.cosmology.omega;phys.matter
    phys.cosmology.omega;phys.darkEnergy
on the principle that different kinds of omega are basically the same
concept and should be distinguished by secondary UCDs...

Along the same lines, we might want to record other kinds of algorithms
like smoothing algorithms applied to data, so
  soft.alg.sim.nbody
might be better than sim.alg.nbody.

In the above I have taken as a general principle: 
 What's most important is the concept, not whether it's simulated or not.
 In many cases what you're simulating has a counterpart in the real universe;
 in those cases the "sim.something" should be a qualifying adjective, not the
 main concept.

So things like sim.obj.particle, which are truly simulation specific,
should stay as Laurie has proposed. I'm not sure about sim.obj.halo, I
guess that's a sim-specific thing, but sim.obj.gal sure isn't.
(Note that the whole issue of a UCD tree for an astronomical object ontology
has recently come up again for the nth time. I'm sure it will eventually
happen despite the naysayers.)

 - Jonathan McDowell