Re: ADMS high level comments

On 2012-10-11, at 12:19 PM, Dave Reynolds wrote:

> (1) My fundamental concern is that it is not clear what makes a Semantic Asset different from any other asset and so why it makes sense for GLD to publish ADMS as well as DCAT.
> 
> The document does try to address this but I, at least, was not convinced.
> 
> The history of metadata tells us that is very hard to distinguish data from metadata. I see no reason why a DataSet in the DCAT sense isn't a Semantic Asset and vice versa.
> 
> So a GLD outsider, who didn't know the political history, might expect ADMS to be essentially a DCAT profile.

There does seem to be a lot of overlap. It'd be nice if ADMS and DCAT used more of the same RDF properties.


> (2) The relationship between the listed classes/properties and actual expression as an RDFS/OWL vocabulary is not sufficiently clear.
> 
> This has been raised before, and I understand the historical reasons behind it, but at some point this will need fixing and *some* clarification is needed before FPWD.
> 
> Apart from the use of informal labels instead of localnames or curis there are properties in here which (correctly) don't exist in the vocabulary itself - notably "id". This makes the document a very misleading guide to the vocabulary.

Yes, I am also wondering if there will be another document to describe the RDF version of ADMS. A GLD outsider would want to know whether a SemanticAsset name ought to be a skos:prefLabel, a rdfs:label or a dcterms:title, for example. It seems like http://www.w3.org/ns/adms is meant to do this, but it doesn't give the RDF for every property in ADMS. Are we meant to proofread that document as well, or will that document only be worked on after we review http://dvcs.w3.org/hg/gld/raw-file/default/adms/index.html?


> (3) The one thing that you do need with semantic assets, that you many not need elsewhere, is information on closure. You need to be able to state that some particular enumeration of codes in a codelist is complete and that a code not listed there is invalid. Is this use case supposed to be supported by ADMS?
> 
> I see that you can represent hierarchical containment of assets through adms:includedAsset but there's nothing about closure or completeness either as guidelines in the document or as a metadata term.

Wouldn't you need some OWL for that? e.g. owl:oneOf for your "code not listed there is invalid" case. ADMS wants to be technology-neutral, so not sure how that sort of axiom is generally described in such documents. I assume users of ADMS could add these sorts of axioms, and that ADMS need not define any itself. It's possible not all users of ADMS will want the same axioms.


> (4) There's a lot of use of the term "file". This seems inappropriate in a W3C spec, especially one about semantic assets. Surely a common case will be things like code lists, represented in SKOS and made available as Linked Data.
> 
> This may be "just" a terminology problem but it is a jarring one.

I searched the document, and all mentions of "file" are near mentions of "distribution." Can you point to a specific problematic case?


> (5) This is a nitpick but it seems odd that translations are distinct SemanticAssets whereas representations are just distinct Distributions. If I represented a schema in RDFS instead of XSLT that's a much bigger change than if I translated it's labels to French.

I would think that translating something like a guideline document or a specification does deserve a distinct SemanticAsset. Are you proposing that a new class be added to represent translations of SemanticAssets, to mirror the additional class that exists to represent distributions of a SemanticAsset?

James

Received on Friday, 12 October 2012 04:53:18 UTC