- From: Toby A Inkster <tai@g5n.co.uk>
- Date: Thu, 10 Apr 2008 18:21:45 +0100
- To: public-rdf-in-xhtml-tf@w3.org
Hello all, First to quickly introduce myself, I'm Toby Inkster. I'm an in-house developer for a UK charity, though my interest in the semantic web is not really related to my day job. I'm working on a GPL'ed "semantics extractor" (for want of a better description) called Cognition. The ultimate aim is to make it into a desktop tool with integrated browser, but for now I'm happy to just parse stuff and export it in interesting formats. As input it supports various microformats (native, non-GRDDL support), RDFa, eRDF, RDF/XML (linked to with <link rel=meta>, or embedded in XHTML directly using namespaces or within HTML comments) and has partial support for GRDDL. Cognition: http://buzzword.org.uk/cognition/ (Should be releasing a new version soon with s/instanceof/typeof/ amongst other changes.) My interests lie within what I like to call the "mixed case semantic web": unifying POSH/microformat data models with the more formal side of the Semantic Web. Making sure that data extracted from one side is available to the other. As an example of the kind of thing I mean, take a look at <http://examples.tobyinkster.co.uk/hcard>, which is a hybrid example of hCard and RDFa. It is correctly parsed by Cognition, and can be output as vCard or RDF/XML. Anyway, I popped my head into the beginning of the task force meeting on IRC this afternoon, to check if it was open to the public, as I had an idea I wanted to contribute for supporting RDFa in HTML (i.e. as opposed to XHTML). The problem as I understand is that xmlns:foo attributes are unusable in HTML as they won't validate. Strictly, they won't validate against the XHTML DOCTYPE either, but we cough and mumble and ignore that because the W3C validator pretends that they're allowed. Anyway, my idea is: RFC 2731 to the rescue! RFC 2731 was a technique proposed by the Dublin Core lot to allow the use of CURIE-like prefixes like "dc:" to be used for HTML <meta> elements. For example, to define the prefix "dc" to point to the current Dublin Core RDF vocab, you could use: <link rel="schema.dc" href="http://purl.org/dc/terms/"> And then the prefix could be used in <meta> elements like: <meta name="dc.creator" content="Toby Inkster"> If this technique for defining prefixes were to be allowed in RDF (though I'd recommend replacing the dot separator with a colon) then RDF in HTML becomes feasible. With RFC 2731 these prefixes are valid document-wide, but it would be theoretically possible to extend RFC 2731 to allow prefixes to have a scope (i.e. equivalent to xmlns attributes on non-root elements) by simply following the general rules of RDFa: <div about="#thisSection"> <span rel="schema:dc" href="http://purl.org/dc/terms/"></span> <h2 property="dc:title">Title of this Section</h2> <!-- ... etc ... --> </div> if desired. However, that might be impractical to implement, because of cases like this: <div about="#thisSection"> <h2 property="dc:title">Title of this Section</h2> <!-- ... etc ... --> <span rel="schema:dc" href="http://purl.org/dc/terms/"></span> </div> so I'd probably suggest restricting this technique to just allow prefixes to be defined through <link> elements in the document head. Anyway, those are my ideas with regards to RDFa in HTML. If anyone has any queries then, let me know either on or off list. By the way, according to the list archives there are mumbles about changing the algorithm for parsing RDFa, particularly with regard to "dangling rels". If this has been decided, could the rdfa-syntax document be updated so that I can catch up? regards -- Toby A Inkster BSc (Hons) ARCS
Received on Thursday, 10 April 2008 17:35:38 UTC