RE: let's specify meaning rather than processing

> How is a processing instruction easier than an attribute?

This came up in responses I had to the piece I did on extending RSS [1]. The
idea was to make non-RDF RSS x.x interpretable as RSS 1.0 (RDF). I'd done it
by adding a (separate-namespaced) attribute:

<ssr:rdf transform="http://ideagraph.net/xmlns/ssr/source/rss2rdf.xsl" />

A couple of people suggested that this should be a PI rather than an
attribute. But this didn't fit with the plan - no processing was necessary,
the stylesheet was only intended to declare the mapping. Ok, most of the
time in practice you probably would have used XSLT -> RDF/XML, but the point
was that the mapping imposed an RDF interpretation on the non-RDF XML, how
you actually parsed it wasn't relevant.

I've a feeling with GRDDL the intention is the other way around - processing
is asked for.

Incidentally a search on the namespace I used suggests that hardly anyone
has used this approach - which in retrospect is understandable : if you
don't want RDF/XML you're not going to be interested anyway, if you do, then
you're likely to bung it through XSLT whether or not there's a declarative
attribute.

Cheers,
Danny.

[1] http://www.xml.com/lpt/a/2003/07/23/extendingrss.html

Received on Monday, 26 January 2004 15:16:15 UTC