Re: ACTION-165: Add a section about publishing metadata to item nr 4

IMHO its worth providing an extension point here to support publishing of
structured machine-readable metadata.  In general, use of catalog-style
metadata does not lead to useful outcomes in a Web-scale - for example the
OGC testbed that recently delivered a resource on a global search called
"Rainfall(mm)" - an example of semantic widening of search, but scarcely a
detailed description of the data, its range, purpose etc.

In particular, there are a number of relevant ontologies that can, today,
provide such extensions ( e.g. PROV, OWL to describe content models, VoiD,
RDF-Datacube - and the emerging SSN).

I think we should make a clear statement about the distinction between
three types of metadata, as found in practice:
1) Descriptive metadata to support human interpretation in a browsing
context and wide/loose search strategies based on language
2) Coded metadata where values are known in some domain (typically this is
just a domain-specific portal) - such as a subject descriptor - and BP here
should be to use dereferencable URIs as such codes
3) structured metadata using RDF graphs based on well-known ontologies
relevant to the aspect being described

The last point I believe needs further debate - is there a canonical way we
can recommend (a BP) of including links to such metadata within the catalog
overview record (i.e. object properties of the dataset description object) ?

I dont believe that the DWBP goes far enough ( to support the SDI
use case.

DWBP4 devolves the responsibility for "structure" metadata to the
"distribution"  - but this has several problems - including the fact that
the platform may not support useful descriptions - e.g. the problem a WFS -
you may get a schema but you cant formulate a query because there is no
information about which attributes are populated, or the range of values
that exist.

As we have a rich set of metadata concerns that are ubiquitous in this
domain, yet the combinatorial explosion of possible alternative
vocabularies makes this a bad place for a free-for-all. shows one
pattern for linking such metadata - where a well-known term is used but the
target object is expected to have extensible machine readable metadata
relevant to its type.

Its going to be a lot of work to have a consistent, useful approach
implemented whichever way we turn, but i feel we could at least provide
some guidance here that in turn makes it easier to express how to use the
other SDWG outputs within the BP context.


On Wed, 25 May 2016 at 08:42 Andrea Perego <>

> Dear colleagues,
> About ACTION-165, I added a draft section in item #4 of the BP Narrative 2:
> In the text, besides references to the relevant (IMO) SDW BPs, I've also
> added the DWBP's ones.
> @Bart, could you please check it out? Also, please feel free to change /
> extend it as you see fit.
> Thanks!
> Andrea

Received on Wednesday, 25 May 2016 00:56:38 UTC