Re: Is it time for eRDF 1.1?

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index]

From: Henri Sivonen <hsivonen@iki.fi>
To: Micah Dubinko <Micah.Dubinko@marklogic.com>
Date: Thu, 7 Aug 2008 10:54:17 +0300

On Jul 30, 2008, at 19:14, Micah Dubinko wrote:

> ERDF is a community spec for embedding RDF triples in valid HTML or  
> XHTML. That is, no additional elements, attributes, or namespaces  
> need to be defined.
>
> http://research.talis.com/2005/erdf/wiki/Main/RdfInHtml
>
> With all the activity around RDFa, now an increasingly-implemented  
> W3C Candidate Rec, I wonder it if’s time to revisit eRDF and perhaps  
> bring it into closer alignment?
>
> http://dubinko.info/blog/2008/07/28/erdf-11-proposal-discussion/
>
> I’m interested in anyone’s thoughts on this.

CCing HTML WG

I think both eRDF and RDFa have the problem they try to make URI-based  
extensibility look shorter by adding a layer of indirection--i.e.  
complexity--in the process. It seems to me that wanting to use URIs as  
identifiers but wanting identifiers to be short is trying to have  
one's cake and eat it too, and the result is complexity and brittleness.

The current draft of HTML5 doesn't have the profile and rev  
attributes, because they are seldom used, profile is used for things  
that the editor of the spec considers bogus and particularly rev is  
often used wrong. On the other hand, HTML5 allows URIs in rel, id,  
class and name (on meta) attribute values (though it requires rel and  
name values to be registered on a wiki).

I think this leads to doing something like eRDF but with full URIs.

Here's a quick HTML5-compatible "eRDF5" trial balloon as a delta spec  
to eRDF:

  * Instead of declaring schemas, use full property URIs where eRDF  
uses prefixed local names. When consuming, assume that a rel, id,  
class or name value that contains a colon is a URI. (Keep prefixing  
RDF classes with '-': e.g. class="-http://xmlns.com/foaf/0.1/Document";.)

  * Instead of requiring a profile, assume that anything that looks  
like eRDF5 is. Consumers simply ignore possible noise triples whose  
property URIs they don't recognize. Where this kind of probabilistic  
approach isn't considered unambiguous enough, any subtree can be  
unambiguously marked as being in the eRDF5 format by adding class='http://purl.org/NET/erdf/profile' 
  on the subtree root.

  * Instead of using <meta name="dc.title"> for backwards compat with  
the dc.foo scheme, become compatible with native metadata and assume  
that HTML5's and HTTP's native metadata maps to Dublin Core assertions  
about the current document. <title> becomes http://purl.org/dc/elements/1.1/title 
, <meta name=author> becomes http://purl.org/dc/elements/1.1/creator,  
<meta name=description> becomes http://purl.org/dc/elements/1.1/description 
, the URI that was dereferenced to fetch the document becomes http://purl.org/dc/elements/1.1/identifier 
, the language of the <body> element (computed with inheritance from  
root element and HTTP) becomes http://purl.org/dc/elements/1.1/ 
language, HTTP Content-Type becomes http://purl.org/dc/elements/1.1/format 
, etc.

  * Reduce the expressiveness of the language by removing the rev  
attribute use.

-- 
Henri Sivonen
hsivonen@iki.fi
http://hsivonen.iki.fi/

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index]