W3C home > Mailing lists > Public > public-lld@w3.org > October 2010

Thoughts on validation / documentation / abstract models

From: Mikael Nilsson <mikael@nilsson.name>
Date: Sat, 23 Oct 2010 21:59:35 +0200
To: public-lld@w3.org
Message-ID: <1287863975.3271.79.camel@daneel>
Hi all!

A follow-up on the discussion on abstract models and application
profiles from yesterday's F2F meeting.

I've been thinking about the discussion about the future of DCAM and the
role of application profiles in documenting / validating metadata
models.

It seems to me there are two perspectives here, both equally valid and
useful, and not even in conflict, really.

The first perspective is the technical. Here, the focus is on syntaxes
and working tools for creating metadata descriptions. We have here, as
an example

* the RDF abstract syntax as a data model
* various RDF syntaxes for metadata exchange
* OWL and RDF Schema as tools for describing semantics
* validation of metadata instances as a technical problem (like XML
Schema)

The other perspective is the documentary view, where the important thing
is to communicate metadata models in an informal way, on the level
of the "domain model" of the Singapore framework, similarly to what
FRBR does. In this perspective:

* UML or OWL can be used to describe a high-level view of a domain model
without reference to any concrete RDF triples. 
* No metadata syntaxes are involved.
* Informal definitions of metadata terms (like "creator") 
* Validation is not interesting, as the model can be implemented in many
different ways (like FRBR). 
 
If I understand the discussion yesterday correctly, this is the value,
the potential, that some see in the DC Abstract model. It is seen a way
of modelling metadata models in an abstract way, without reference to a
concrete metadata language.

I understand the value in the documentary view, but must consider it to
be a misunderstanding of the role of the DCAM. DCAM was always about
specifying metadata in a way that is bound one-to-one to concrete
metadata instances (unlike metadata specified like FRBR, which can be
expressed in many ways even in RDF).

So, my conclusion is that there is room for a metadata modelling
language of the kind envisioned above. I guess the term "abstract model"
might lead you to think DCAM is such a language, but IMHO it's not, not
at all. 

DCAM is, in fact, very close to the RDF abstract syntax in its purpose.
Any notion that metadata described using DCAM is portable to a variety
of metadata syntaxes is just plain wrong. DCAM is much to detailed and
technical for that.

However, perhaps there is room for such a "metadata modelling" language?
I'm thinking about something like an UML profile designed specifically
for metadata.

Such a language would be completely orthogonal to an abstract syntax for
metadata like RDF (which is indeed an "implementation"), and also to the
notion of validation - such a model is too high-level to lend itself to
instance validation.

Perhaps confirming the existence of these two possibilities is a way
forward here? We just need to make clear that no, DCAM is *way* to
concrete to be useful for this. 

And validation belongs to the technical perspective together with DCAM
and RDF, while application profile *documentation* would perhaps be
better handled by a metadata modelling language?

/Mikael
Received on Saturday, 23 October 2010 20:00:12 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Saturday, 23 October 2010 20:00:13 GMT