SSA UTYPES

[Doug Tody]

Hi again Alberto -

> On Tue, 19 Dec 2006, Alberto Micol wrote:
> 
> > On Dec 19, 2006, at 19:07, Doug Tody wrote:
> >> The motivation for this is that we would like to use most of the
> >> same metadata for other DAL interfaces (at this level very little is
> >> specific to spectra),
> >
> > Yes, though I would distinguish between the current phase (SSA V1)
> > and a subsequent phase (let's call it V2 for the sake of a name).
> >
> > If I get the picture right, the main goal of V2 should be
> > to unify SSA and SIA by using a common underneath data model
> > (probably the characterisation dm), allowing thus common metadata
> > to describe n-dimensional data, being them spectra, images, data cubes, etc.
> >
> > In V1 if I'm not wrong, the agreement is that SSA is based on the Spectrum DM 
> > 1.0.
> > The idea being to reach fast SSA1.0, leaving to the next version the role
> > of unifying things.
> >
> > Basically I see V1 and V2 as well defined and intrinsically different.
> >
> > Instead, reading SSA0.97 I get the impression that SSA1.0 will
> > be somewhere in between V1 and V2; and there is where my difficulty
> > originates. To state it clearly I think that SSA1.0 should
> > use Spectrum DM 1.0 as is (where possible of course), and not
> > deviate from it, as instead proposed in that 3.3.6 paragraph,
> > or as alluded in the Appendix 1 "Mixed Service".

The Spectrum data model was developed as part of SSA.  The SSA
and Spectrum data models are identical (ignoring minor namespace
differences which do not affect the semantics) except for the addition
of some extra things in SSA (query metadata, association metadata,
access metadata) which are not relevant to Spectrum.

It is to be expected that when we come out with SIA V2 and TAP, and
also support for time series and SEDs, that there will be another
version of these data models and probably an update to SSA, however
I would not expect it to change a great deal at this point.

Although Char is a major part of the data model, that is not all there
is to generic dataset metadata.  The other models such as Dataset,
DataID, Curation, Target, etc. are equally important.  As you say,
this "generic dataset metadata" is intended to be the same for all
datasets, e.g., spectra, images, cubes, etc.

Which brings up an important point - what we are really creating are
component data models.  If we think in terms of component data models,
then the SSA query response and Spectrum are in a sense a packaging of
these components.  The fundamental models are the same, even though the
mixture of things packaged together may change.  I think components should
be the focus of our data model re-use in different contexts.


> >> plus at this level we prefer to have the UTYPEs
> >> rooted in each component data model.  Hence, a UTYPE is something like
> >> "ssa:Target.Name".
> >
> > So, the answer to my question was that it is an ssa name space then?!
> >
> > Thanks,
> >
> > Alberto

Yes.

 	- Doug