re: Pre-meeting UCD question

Hi all,

Roy asked:

> (2) Perhaps we can start a discussion by asking for requirements in trhe
> following question:
> *****  WHAT ARE UCD USED FOR -- now and future *****

First up, I'm new, so please forgive (and correct) any misconceptions on my part :-) Anyway, here is my 2 cents worth. I'm not sure if I'll be able to make the meeting, but we're hoping to have someone from Aus-VO there...

> Ref 2: UCD Status and Perspectives, S. Derriere > http://www.ivoa.net/internal/IVOA/IvoaUCD/UCD-status0203.pdf

On the top of page 3 of this document, I am told that UCDs are not:
- accurate definitions,

• expressing physical dimensionality, or
• tied to units.

Why not? I would imagine that UCDs would be most useful if they were *all* of the above.

Accurate definitions: if they are not accurate, what good are they? Why would I want to search a bunch of UCDs to find one that *might* measure what I'm looking for, rather than one that *does*? I imagine an accurate definition would be needed for this.

Expressing physical dimensionality: why shouldn't they? This is a simple concept, and one that certainly could assist with at least figuring out if two apparently different UCDs can actually be compared. It's certainly not in and of its own a powerful concept, but it's perfectly possible that I might want to look for any columns which measure a length of time... On the other hand, physical dimensionality can often be derived from the units, so this is not a critical inclusion.

Tied to units: again, I'm not clear on why they shouldn't be in the UCD. If we don't store units in the UCD, they have to be stored somewhere else, and then the column values are useless without *both* the UCD and this other bit of information.

Overall I am in favour of having all information which describes a single measurement stored in the one place. The UCD seems like a good candidate to be that place.

More generally, I felt that one way to categorize UCDs would be as:
- instrumental UCDs

• measurement UCDs
• derived UCDs in the sense that any particular column is generally going to be either, respectively:
• a setting you made on your telescope or in your theoretical software (eg. a table of frequencies corresponding to each channel of a spectrometer, or the number of particles in your simulation, ...)
• a measurement you actually made (eg. the flux density at each channel of your spectrometer, the energy of each of your simulation particles)
• a derived quantity, such as the line-integrated flux of your spectrum, or the total energy of your particles Then measurement UCDs would include "links" to the minimal set of instrumental UCDs needed to repeat the measurement, and derived UCDs would include "links" to the minimal set of measurement UCDs (and implicitly instrumental UCDs) to repeat the obtaining of the derived value. I think a scheme like this would enable some pretty advanced searching, such as finding catalogues (or more generally sources of data) who measured the same thing as you but in a different way. Just one idea really.

Maybe I have sidetracked on to the data model, please let me know if I have. Apologies if all the above has been said and discussed in the past. Interested to hear what others think of the above.

David. Received on 2003-04-07Z04:28:33