On 26 May 2009, at 08:14, Henri Sivonen wrote:
In a reasonable layering, the RDFa processor is layered on top of an HTML parser, in which case the RDFa parser can't in the general case tell if the input would have been well-formed if processed as XML. It seems like a bad idea to define RDFa processing in terms of the source bytes instead of defining it in terms of the output of the HTML parsing algorithm.
Indeed. That is a good argument in favour of solution #3 from my previous message.
Most current RDFa parsers do seem to sit on top of a separate (X)(HT) ML parser, just looking at a DOM tree which can be fairly reliably serialised to XML, so this solution probably has the smallest implementation cost for them. The exception is XSLT-based implementations; but, given that XSLT processors can't operate on HTML by definition, this seems to be a moot point.
-- Toby A Inkster <mailto:m...@tobyinkster.co.uk> <http://tobyinkster.co.uk>
