RE: RDF Issue: mime-types-for-rdf-docs

>I am only half agreeing with you, Danny.

'Only'? I think that's a record ;-)

>1. About the parser recognizing the type
>
>If I push your argument one step further, why wouldn't any XML parser be
>able to recognize that a document is RDF or SVG, and then behave
>correctly ?
>The first part of the assertion (recognizing) is technically feasible
>(just check the namespace of the root element); but the second one
>(behaving) is not.
>
>As well as you need a specialized parser to handle SVG documents
>(meaning, a parser able to actually draw the graphics), you need
>specialiazed RDF parsers to handle picx, xpackage or annotea content.
>
>Hence it looks consistent to me to create specialized mime-types for
>those. It does not mean that a more general parser can not do *anything*
>with these types: indeed, an XML parser won't be totally useless with
>some application/pics+rdf+xml content. This is the whole point of the
>"+xml" suffix.

Splitting not only hairs but functionality too - an initial parser+router
could check the root element and forward as appropriate, there is no reason
to expect this part of the process to be much concerned with the content,
beyond the root (for convenience it could I suppose pass a DOM tree or an
RDF graph to subsequent stages, but that's just implementation).

>2. About mixing namespaces
>
>It is true that, with RDF even more than XML, content types can be
>intermingled: an XHTML document may contain embeded SVG, but this
>"embeded" part can quite easily be isolated and redirected to the
>appropriate specialize parser.

Quite, so the idea isn't that alien.

 In RDF, isolating parts related to a
>particular namespace may be a puzzle.

Hmm - I'll have to think about this.

>Anyway, I think that there are a lot of situations where it is possible
>(and easier) to see an RDF graph as
> - conforming to a particular schema (say, the "primary" schema),
> - and *enriched* by additionnal information from other schemas (say,
>"additionnal" schema)

Indeed, but it's early days - it may be that real mixup graphs turn out to
be very useful, I would hesitate to design around what's easy *now*.
Remember y2k?

>even though the "primariness" of the schema is not intrinsic to the
>graph, but just determined by a contextual choice (for example, to
>conform with the types acceptable by the client, like in HTTP).

That sounds perfectly reasonable - let the client (agent) decide.

Cheers,
Danny.

Received on Tuesday, 9 April 2002 16:09:30 UTC