UCDS vs DM - Peacekeeping

[Tom McGlynn]

Martin,

So far I've been a non-combatant in this war, but as you may know I had some
reservations about the UCD2 framework presented at the interoperability 
meeting.  Over the
past two days I have been writing a revised proposal.  This proposal 
includes a whole
new discussion of the interaction between UCDs and table grouping 
constructs that I feel may play a
major role in mediating among UCDs, data models and the DAL.  I'll be 
publishing this within a day
or two, but one key concept is the UCDTree which shows the UCDs of a 
table in a structured way.  It is
my hope that methods of abstract data models can be translated into 
actions on data by straightforward
analysis of UCDTree.   I expect to be send out the proposal today or 
tomorrow after I have a chance to check
that what I've written during the flight back makes sense when I read it 
in a non-sleep-deprived state.

    Regards,
    Tom

martin hill wrote:

>I've been trying to sort out in my own head the differences between UCD2s and 
>data models.  Particularly as one doesn't seem to work entirely without the 
>other.  So donning my UN peacekeepers hat (which in the British case is a 
>tatty gardeners hat in a rather trendy camouflage, not a kevlar helmet):
>
>It strikes me that data models are about structure, and UCDs about describing 
>elements in that structure. 
>
>Now it is probably possible that the way data models are defined could include 
>naming elements to define what they mean.  I suggest that these should be (or 
>include as attributes) UCDs, at the very least so that we can compare data 
>items that have been formally modelled with those that haven't.
>
>For example, we can say (simplistically) that a coordinate is an RA, DEC, 
>error, and refers to some co-ordinate frame, and might look like this in XML:
>
><CatalogueObject ucd2="sky.galaxy">
>  <WorldCoordinate ucd2="pos" frame="J2000">
>     <RA ucd2="pos.wcs.ra">
>        <VALUE units="degrees">42</VALUE>
>        <ERROR ucd2="error" units="arcSeconds">23</ERROR>
>     </RA>
>     <DEC ucd2="pos.wcs.dec">
>        <VALUE units="degrees">42</VALUE>
>        <ERROR ucd2="error" units="arcSeconds">23</ERROR>
>     </DEC>
>  </WorldCoordinate>
>
>  <Brightness ucd2="phot" type="SomeObscureOptical">
>      <Magnitude ucd2="phot.mag.vega">72.3</Magnitude>
>      <Error ucd2="error">1</Error>
>  </Brightness>
>
>  etc
></CatalogueObject>
>
>Now this is a horribly simple example (sorry about the mixed-up case 
>conventions) - how do people feel about it? It means that we should avoid 
>trying to describe structure/context in UCDs (which has the potential of 
>making them horribly long and complicated) and gives us an immediately useful 
>way of giving wider meaning to our data structures.
>
>It kind of implies that we then have a method for appending our UCDs up a data 
>model tree if we need to get more context for them.  Thus we don't have to 
>have src.galaxy;phot.mag.ObscureOptical;error *as a defined UCD*.  Instead 
>such strings are constructed out of individual UCDs as required by the program 
>that is investigating the data.
>
>It also means that UCDs don't have to be specific (which the UCD group are 
>avoiding cos it's a horrible task, small wonder) and yet I as a developer can 
>assemble specifics for doing cross comparisons.
>
>I've only had a pint and it still seems a good idea. It was a big pint though.
>
>
>  
>