Comments on DXWG UCR document

Dear Ana

thank you for your comments on 20 Jan 2018, and apologies for the delay in
response.

There has been some wider discussion around some of the points and as part
of the UCR editing team I needed to confirm my understanding of the group
direction.

Regarding the first point we have now agreed on a working definition of a
profile, and have deferred any definition of subclasses of this concept.

The working definition is:

*"A named set of constraints on one or more identified base specifications,
including the identification of any implementing subclasses of datatypes,
semantic interpretations, vocabularies, options and parameters of those
base specifications necessary to accomplish a particular function."*[1]

[1] <https://www.w3.org/2018/02/06-dxwg-minutes>(
https://www.w3.org/2018/02/06-dxwg-minutes)


I think this definition covers the cases addressed in
http://www.ariadne.ac.uk/issue25/app-profiles


With regard to the proposed additional requirements, we have discussed but
find ourselves needing greater clarity of the intention.  Note that all
requirements should have supporting Use Cases, and Use Cases in turn are
best supported with examples and "real world" usage links.


What follows are some comments based on my own personal understanding (or
misunderstanding) of these cases:

I think "Multidimensional Metadata Profiles" is a more generalised
description that is perhaps implicit in the definition above. The specific
requirement for a "canonical view" is stronger than we have yet adopted,
and if a case were to be made for this it can be discussed.


Regarding a range of a controlled vocabulary - i did some experiments with
this with RDF-QB (datacube) - the only model in the W3C canon I know of
that explicitly binds controlled vocabularies (using SKOS) to structural
data elements without requiring such vocabularies to be imported
ontologies. It allows a rdfs:range to be defined, which means that ranges
can be defined for hierarchical vocabularies based on skos:Concepts with
additional domain types - e.g.
Kingdom/Phylum/Class/Order/Family/Genus/Species in biological taxa. A
filter is not possible however, so keen to see an example of how this might
be done.

Grouping terms form one or more vocabularies is trickier. Perhaps this is
an out-of-band issue - the "effective vocabulary" that results may need to
be deployed as a separate artefact and the profile mechanism kept simple.


Thanks again for your input, provide any more details you feel can help
progress these ideas and please stay engaged as we progress.


Rob Atkinson