[DM] Plea for constructive progress (Was: Re: [QUANTITY] Plea for pragmatism)

[Doug Tody]

> The difference in our approach is that I start from the user requirements
> (read: what an astronomer wants from the DM WG), and
> from there I will get to the software and architectural requirements,
> and not viceversa. The CS requirements come only later in the process.

At the astronomer level perhaps the requirements are for the VO overall
(VO usage scenarious and so forth) and we have to consider architecture
and requirements analysis to determine what is needed at the level of
a particular element such as the data models.  To some extent we have
already done this over the past 2-3 years and our development is already
organized with a certain system architecture in mind.

VO data access is data model based so we have some immediate use cases
for data models from DAL:

    o	Simple data models for image (SIA) and spectrum/SED (SSA).  SIA 1.0
	largely side-steps this issue, but SSA will need a more explicit
	spectral/SED data model.  We already have a good start on these.

    o	How to characterize datasets such as SIA and SSA, i.e., the coverage
	issue, component data models, etc.  What are the components?
	This is tricky, for example time of observation is actually
	a bandpass, spatial resolution is related to spatial bandpass,
	limiting flux could be an attribute of some sort of "flux bandpass",
	and so forth.

    o	How to represent the component data models in a DAL query response,
	e.g. VOTable, to summarize the characteristics of a dataset which
	is a candidate for access.  This gets into issues such as UTYPE and
	data model representation in XML, UCD2, grouping constructs, etc.

    o	How to use dataset characteristics to refine a query.  Any dataset
	characteristic which we model will be used both to describe a
	dataset, and to refine queries looking for such datasets (e.g.,
	can we reduce spectral bandpass to a filter name or coarse bandpass
	ID such as optical/ir/radio etc.).

    o	How to represent actual datasets in various data formats such
	as XML.  We need to map external data into our VO data models
	and represent such virtual data in XML and (to a lesser extent)
	other formats.	This is closely related to the query response, but
	some actual data is passed as well.  DAL will return data in these
	formats, e.g., a 1D spectrum or an image cutout.  Functions will
	need to be able to operate directly on data in this form.

Doug