[QUANTITY] Plea for pragmatism

[David Berry]

Pierre,

> Data containers ARE NOT restricted to QUANTITY, at least for some of us.
> That a major point on which we have to agree before continuing modelling,
> I agree it is part in a certain manner of high level modelling decision.
> If we agree to separate these two domains, I think that we have made a great
> step forward (hoping we are not just standing in front of a pit).
> Can we try to achieve something very simple, like the simple quantity proposed
> by Pat (off the list), before trying to use our skill and energy
> on complex issues like "Universal Data Container".

I agree that "Data Container" is not a good name for what I described
since a data container should be a purely structral thing like an "array"
or a "tree", etc. The phrase "Data Container" may perhaps more rightly
apply to things like my "DATA component" rather than the whole ensemble of
components. Maybe I *should* just stick with QUANTITY instead of "Data
Container".

> > Having said that, I think it would be useful to have a simple overview of
> > what a candidate data container may look like, just so that we can check
> > if it is able to meet the requirements which emerge as we develope the
> > OBSERVATION model, and change it as needed. I'm thinking here of something
> > as simple as the following:
> >
>
> Once we have to tackle this subject, I think I would agree with most
> of the points below, except QUALITY, and add COVERAGE, COMPLETNESS
> (lacunarity) at least.

I was thinking of COVERAGE and COMPLETENESS as a components of the
OBSERVATION rather than the QUANTITY. You may want to use a QUANTITY to
define the coverage. For instance, the spectral coverage of an observation
could be specified by a 1D QUANTITY holding a bandpass.

> > QUALITY - gives a set of flags and/or enumerated values for each value in
> > the DATA component.
>
> This is a very complex information, very diffucult to attribute, which can
> evolve with time, depend on the subject studied...
> One data set unusable for a certain study, can be perfect (if not better)
> for another one.

Again, this is not suppose to characterise the overall usefullnes of the
observation for some purpose, but just simply to assign specified flags to
individual data values (e.g. "this value is saturated", "this value
is vignetted", etc ). Characterising the usefullness of the whole
observation would not be the task of a QUANTITY.


> > WCS - Contains a collection of world coordinate systems in which
> > positions within the DATA component can be described, together with
> > Mappings which describe how to transform positions between different world
> > coordinate systems.
> >
>
> Reference frame is more general and certainly more appropriate?

The idea is that the position of a data value can generally be described
in terms of several different coordinate systems, not just one "reference
frame". For instance, for a CCD image of the sky, a position may be given
in pixel coordinates, or focal plane coordinates, or sky coordinates. The
WCS component would contain descriptions of all of these coordinate
systems, together with Mappings which allow a position to be transformed
between any two coordinate systems. This could be done in a "toolkit"
manner to allow the system to be extended easily to cover data (e.g. solar
or STP data) which is more complex than a simple CCD image of the sky.
See:

http://www.ivoa.net/internal/IVOA/InterOpMay2003DataModel/Structuring-WCS.ppt
(or, if you object to powerpoint,
http://axp0.ast.man.ac.uk/~dsb/ast/IVOA_WCS_talk.html)

> but I think we must concentrate on the domain definition, (the global
> view your asking for [OBSERVATIONS]) and we can at the same time
> try to model a small well delimited concept to exchange (hopefully
> efficiently) small amount of data (atomic value or small arrays).

Agreed.

David