Data Models 2: Minutes of questions and discussion, by JCM

Mireille Louys presented the latest work on the Characterization model
and Francois Bonnarel presented a possible serialization. The suggestion
that Accuracy be made part of Characterization rather than part of the
data generated some discussion. Jonathan argued that the errors should
be close to the data while Doug pointed out that different levels of
noise model may be useful. It wasn't entirely clear what
the definition of Accuracy was in this model.
 Another issue was whether complicated data
like exposure maps should be in characterization or provenance.
We reiterated that provenance should contain data with instrumental
signature while characterization should have the stuff that is 'pure'
physics. Doug's view is that detailed calibration doesn't belong
in characterization, while Jonathan believes that it ultimately
does, although of course not for the query-response aspects used in
the SSAP protocol.

Francois' presentation of the proposed XML for Char. included
an element time defined by restriction on a particular UCD value,
and we asked for some examples to clarify this in both votable
and XML instance. FB commented that this construct
was partly due to the removal of the Frame object
from Brian's version of the Quantity model. It was suggested (GL?)
that instead of inheriting things from the quantity model
it should be reused. Arnold warned that restrictions mixed with
substitution groups can lead to problems in the resulting code.

Igor presented on 3D IFU data. I note that the concepts of
line spread function (as a function of instr.x, instr.y, lambda) and spatial
PSF (as a function of mirror.x, mirror.y? different from instr.x instr.y
in the case of image slicers?) came up as something we may need to handle.
(they fit in Resolution.Spectral and Resolution.Spatial of course).
We should review the Euro3D format and see how it maps to the SED model.
The issue of curving/optical distortions on the IFU spectrum
was raised. It was commented that the data quality flags recorded different
parameters of the instrumental process and so had a stronger link to
Provenance than Characterization.

In a general discussion on 3D data it was felt that two main approaches
- a 3D data cube x,y,lambda, and an aggregate of multiple 1D spectra
each tagged with a Support.Spatial region description - covered all the
main cases, and that we could define a mapping between the two
representations. In the latter case (aggregate) of course you avoid
rebinning in the spatial domain, which makes people happy. We noted that
in radio, the former case, simple image-spectral cubes, are common. In
the former case you need to deal with the WCS, which we haven't
addressed in detail so far in the IVOA. Characterization doesn't include
WCS, and this may be good because it can apply to non-binned data for
which WCS does not apply. 

We briefly discussed the roadmap for characterization, including a revised
XSD and more instance examples. JCM agreed to provide an event list.

Norman presented work on OWL reasoning engines. The idea is to make
different ontologies interoperable by mapping only a few fields in
them and deducing further compatibilities. JCM expressed concern that
in complicated cases, only lowest leaves on the tree might have
counterparts.

Gerard berated us for not fully following UML syntax and diagram
rules, and proposed an IVOA standard subset syntax that we should
all follow, including bindings to XSD, Java, databases. One problem
he commented on in his scheme was for cases where an object
relationship might point outside a document, e.g. a class might
refer to a human, but the object describing the human might be
in some external resource. Importing via Xinclude (F) or
index pointers for external instances (AR) were suggested.

Gerard pointed out a problem with using element references is
the need to have a top element in the document, limiting reuse.
He will compare his STC schema binding to Arnold's and report.
He recommended avoiding restrictions and using extensions.

 - Jonathan