Representing data

[Doug Tody]

>From Martin:
> So during the meanwhilst - can anyone suggest an email list for discussing
> 'data representations'?  This should include my favourite bee-in-the-bonnet
> topic, inter-service message formats.  Should we start a new one, or perhaps
> use interop at ivoa.net?

I suggest using dal at ivoa.net for such discussions.  Within the NVO at
least, data representation is considered part of data access, since a
DAL service has to return data in some form.  The major elements of a
DAL service are the query and the data access method, which returns some
virtual (computed on demand) data object conformant to the VO data models.

Ideally we define data models in the abstract, perhaps including a
verification schema (this is the task of the DM working group), and can
represent data in various ways, e.g., VOTable or some other XML format,
FITS, or a combination of XML (for metadata) and FITS (for binary data).
The data model should be the same regardless of the data representation,
allowing software to be written for a particular data object or component
data model to be usable regardless of the access protocol and data
representation used.  We may use slightly different representations for
small datasets and for bulk data.

We have an immediate need to do this for spectral and time series data for
the SSA interface.  We can use SSA as a test case to experiment with new
approaches to data representation, e.g., by extending VOTable to allow
datasets to be composed from data model components, each with their own
namespace and schema, combined with storage elements such as table.

	- Doug