New UCD1+ for SIA protocol
dtody at aoc.nrao.edu
Thu Mar 25 07:59:16 PST 2004
Taking this one step further: if we start inventing UCDs which are nothing
more than the names of data model attributes, this may indicate that
something is wrong. Perhaps what we should be doing is using UTYPE
to define the data model attribute or interface element (if there is one),
and UCD to say something about the quantity given as the value.
For example, if the UTYPE is something like image.scale, then the UCD
might define the type of physical quantity including units. Currently
we try to fix the quantity/units, but there are cases where this needs
to vary even though we are talking about a single data model attribute.
On Thu, 25 Mar 2004, Doug Tody wrote:
> On Thu, 25 Mar 2004, Robert Hanisch wrote:
> > I just want to add my support for Alberto's objection to instr.scale -- I
> > think this could be very confusing. The scale here is not an attribute of
> > the instrument, but of the image. Glancing quickly through the UCD-UCD1+
> > mapping, however, I do not see a UCD that fits very well.
> I also agree, this is the image scale not an instrument attribute.
> A larger point here is that in the process of normalizing the UCDS in
> SIA V1.1, we also plan to introduce UTYPE. The recommendation for folks
> writing software interfaces to interpret SIA tables will be to use UTYPE
> to identify the formally defined elements of the interface. This is very
> straightforward whereas UCD is not; we can evolve UCD independently and
> our interfaces will not break. Hence it is up to the UCD folks to assign
> the best UCDs for these fields, and if we have to adjust or generalize them
> later we will be able do so.
> In general I think UCD should be used for hard things like inference,
> semantic associations, smart queries, etc. UTYPE is much simpler and is
> what we need to identify interface elements in a straightforward manner.
> - Doug
More information about the ucd