Handling data cubes in VO

[Doug Tody]

Hi Francois -

I should say more about the following comment (made in reference to the
CGPS/Aladin cube viewer):

> I have nothing against Doug's solution, but I would not like the
> philosophy of what we have implemented being discarded as "NonVOStandard"
> without being considered and discussed.

I agree with the philosophy and overall approach, and very much like the
way the cube is represented as a scientifically comprehensible object
in the Aladin visualization.  The same goes for being able to view a
DAL query response in "tree" form, with various sortings to reveal the
organization of the data.  These capabilities, which have been demonstrated
by the various CDS and Euro-VO prototypes, are absolutely first rate and
serve to demonstrate the kind of functionality we need to evolve the IVOA
standards sufficiently to support.

The overall approach for the CGPS/Aladin spectral cube viewer is to break
the interaction into a discovery query, which describes the cube and its
related views and access methods, and the actual access, which permits
multiple successive accesses to get portions of the cube following a
single discovery query.  I agree that this approach is what we need for
precise data access at this level.

The problem is that in the current version, DAL is being bypassed for
the actual data access.  This will cause problems as I alluded to in
my last mail, e.g., the dataset metadata (for a subset such as a slice)
cannot be retrieved in a standard form, or retrieved separately from the
data access, and other functionality planned for the access interface, 
such as authentication and data staging capabilties, will not be possible
since the standard access methods are not used.

So, the main things we need to do next are to evolve the access methods to
the point where they provide the precise control required for operations
such as subsetting a cube (or simpler 2D image), and define more elaborate
metadata to describe complex aggregate datasets which are built up from
simpler components (such as associating a preview with a data image, etc.).

The end result should support something very much like the CGPS/Aladin
integration, but will be general enough for general ND image access to
support also 2D access, integration with true 3D cube viewer tools (several
of which already exist), and so on.

I think a similar approach is required at the top level to describe
complex spectral data, for example an IFU/MOS aggregate, however this
data is sufficiently different at the level of the actual data samples
that a different interface (SSA) is required.

 	- Doug