« 3DVisA Bulletin | Main | "Beyond control" at ALT-C 2007 »

September 03, 2007

DCMI Architecture Forum at DC-2007

I spent the second half of Tuesday afternoon at DC-2007 in a meeting of the DCMI Architecture Forum, chaired by Mikael Nilsson (Mikael had a very busy day!).

The session started with some further discussion of the draft Description Set Profile model, which had been presented in the plenary session immediately preceding the meeting. Based on the experience of applying the model to the case of the ePrints DCAP, I presented a couple of issues related to the nature of constraints supported by the DSP model, pointing out that it may be the case that there are some constraints that are probably quite widely used within existing DC application profiles that it is not possible to express within the proposed model (see the last slide of my earlier presentation).

The second area of discussion was the question of support for a "modular" approach to the DSP, i.e. the notion that one DSP might be able to "import" or "include" either the whole or some part of one or more other DSPs. I recall that Andy, Mikael, Tom Baker and I pondered this when we met up in Barcelona back in March when we were just starting to think about this work, and I think it soon became clear then that such an approach introduced a considerable amount of complexity, e.g. questions of the relative priorities given to constraints in the two DSPs, whether one DSP could extend or override constraints in the other, and so on. While support for modularity would certainly increase the flexibility of the DSP approach, I think it may be that, in the short term at least, this operates only at the level of "copy and pasting" a defined set of description templates or statement templates into a new DSP.

There was also some discussion of whether DCMI should formulate something like a "Qualified Dublin Core" DSP, referencing the set of terms constituting the DCMI metadata vocabularies (or some subset of them). I must admit I have mixed feelings about this approach: I think one of the problems that DCMI has faced is a tendency for implementers to approach the development of a DCAP by taking the complete set of terms constituting the DCMI metadata vocabularies as their starting point, and adopting something of a checklist approach ("Yes, we'll use A; no, we won't use B; not sure about C"), based on the premise that the use of that set of terms constitutes some sort of "generic DSP", rather than starting from what specific requirements it is necessary to address in the contect of their application. As Mikael pointed out in the meeting, the full DCMI term set now includes some properties that are specific to the description of collections, and some to the description of bibliographic resources - so a DSP that referenced all of those within a single description template would, by definition, be dealing with the description of a peculiarly narrow set of resources. The model presented by Tom which I referred to a couple of posts back - a model which by the end of the conference was being referred to as the "Singapore Framework" - seeks to steer implementers away from this approach, and to root the development of a DCAP firmly in an analysis of the operations that the metadata instances are to support.

We moved on to discuss - briefly - the current proposals for an XML format for representing a DC description set, particularly from the perspective of the sort of structural constraints described by the DSP model. There are currently two working drafts - DC-XML-Full, DC-XML-Minunder discussion on the DC Architecture mailing list. I think the three-way relationships between the characteristics of the XML format, the DSP constraints and XML schema language are potentially quite complex: given format A, it may be relatively easy to capture DSP constraint X using schema language P, but difficult to capture the same constraint using schema language Q; and so on. I don't think any clear conclusions were reached during the meeting -  though there was a strong feeling expressed that we should avoid creating a format (or formats) which allow too much "flexibility" by offering many different alternative syntactic representations of a single construct within the description set abstract information structure (a criticism that has sometimes been levelled, for example, at the RDF/XML syntax for RDF).

At the time, I was slightly disappointed that we didn't get as far as focusing in on making some practical recommendations for how to proceed in some areas, though it was probably inevitable in the circumstances - and it was certainly useful to have the discussions which did take place. Later in the week, a small group of us managed to get together for a chat with Ivan Herman, the W3C Semantic Web Activity Lead, and Alistair Miles of STFC Rutherford Appleton Laboratory, who is the co-editor of the SKOS specifications, and we talked through some of the low level technical issues currently exercising the DCMI Architecture Forum - including those around GRDDL, RDFa and the updating of the DCMI recommendation for expressing DC metadata using X/HTML meta and link elements (again, an initial draft has been circulated). Thanks to Ivan and Alistair, I think we came away with a better picture of how to move things forward in the short term, which hopefully will be translated into some action on the DC Architecture mailing list in the near future.


TrackBack URL for this entry:

Listed below are links to weblogs that reference DCMI Architecture Forum at DC-2007:


I just noticed that Ivan has a post giving his views of the week over at


The comments to this entry are closed.



eFoundations is powered by TypePad