[QUANTITY] Requirements and apology

Doug Tody dtody at nrao.edu
Thu Oct 30 11:25:12 PST 2003


Here is my view of the hierarchy and terminology involved here:

    o	Data is organized into data collections, each of which consists
	of some number of similar datasets.

    o	Individual datasets can be logically associated to form groupings
	such as observing runs, observing programs, and so forth.

    o	Each dataset is something like an "image" or possibly an
	"observation" - although dataset is more general than observation
	since we have composite mosaics and so forth, and an actual
	observation might be represented as multiple datasets.

    o	A dataset is an aggregation and association of 1) general resource
	type metadata; 2) component data models, used to describe coverage,
	data characterization, data quality, and so forth; and 3) data
	instance components of some sort, used to store data values if
	any, error arrays, and other numerical data.  At the level of a
	dataset we can model "images", "spectra", and so forth.

	To represent real-world datasets we will probably require both
	standard elements of types 1-3 above, plus extensions that are
	added by the individual data provider to include additional
	collection-specific information which is not part of the core model.

    o	Quantity and UCD define fundamental physical quantities which
	can be understood and used stand-alone, in data models, queries,
	and so forth.

Exactly where catalogs fit into this picture is not clear.  They are related
generally do not require as much structure.  On the other hand, there
is not a whole lot of difference between an object in a source catalog,
and a 1D spectrum, which might also be stored as a record in a catalog.

	- Doug



On Thu, 30 Oct 2003, David Berry wrote:

> Doug,
> 
> > If we are talking about data values, WCS, data quality, etc., then we
> > are modeling an entire dataset, not a quantity.  We need to address both
> > issues but they are at opposite ends of the spectrum.  Quantity needs
> > to be a low level concept (about the same level as a UCD) or it is being
> > confused with something else.
> 
> I would be happy to call it a "dataset" instead of a "quantity". If a
> Quantity was simply a combination of a UCD and a Units object, then
> a "dataset" would come in between Component and Quantity in terms of
> the Container/Component/Quantity hierarchy you outline in a recent
> message - Component would contain zero, one or more DataSets; and DataSet
> would contain zero or one Quantities.
> 
> David
> 
> 



More information about the dm mailing list