Re: Semantic Web User Agent Conformance

On Nov 23, 2007 10:14 AM, Renato Golin <renato@ebi.ac.uk> wrote:

> The only meaningful place I can see you could put magic numbers
> for all kinds of RDF is in the RDF header itself. such as:
>
> <rdf:RDF (...) magic_constant_1="42" (...)>

I'm not sure that that would ever be required, and it'd be an
extension to RDF/XML anyway. In the case of RDF Stylesheets I'm quite
sure you could get away with magic triples rather than using some kind
of attribute, though as I say I'm not entirely sure what form it
should take yet. It's mainly academic at this point, though I did try
my hand at making a language with mergeable stylesheets that can be
applied to streaming input. Cuts down on the network accesses.

Incidentally I noticed the other day that the application/rdf+xml RFC
is incorrect in its magic numbers section:

[[[
         Magic number(s): none

            Although no byte sequences can be counted on to consistently
            identify RDF, RDF documents will have the sequence
            "http://www.w3.org/1999/02/22-rdf-syntax-ns#" to identify
            the RDF namespace. This will usually be towards the top of
            the document.
]]] - http://www.ietf.org/rfc/rfc3870.txt

Actually the following document is valid RDF/XML but doesn't use the
RDF namespace anywhere:

<SomeClass xmlns="http://example.org/#">
   <prop>Hello</prop>
</SomeClass>

Of course RDF/XML documents served using one of the generic XML media
types have to use <rdf:RDF>, or else there must be a hint given to the
parser.

> I always like to give the Mozilla/IE/Firefox history as an example to show
> that conformity, when compulsory, makes anyone's life much easier.

Heh, indeed. I don't mind walking on that thin blue line between
description and prescription.

> > that makes it hard when you want to merge mergeable stylesheets.
>
> You'll end up with a cascading rdf:type not necessarily compatible, I see...

Yeah, so just to spell it out if you have the following three documents:

<> a :Stylesheet .
- one.rdf

<> a :Stylesheet .
- two.rdf

<> a :Stylesheet .
- three.rdf

And then you do an RDF merge of them, in say cwm or whatever:

cwm one.rdf two.rdf three.rdf > output.rdf

Then output.rdf ends up being:

<one.rdf> a :Stylesheet .
<two.rdf> a :Stylesheet .
<three.rdf> a :Stylesheet .

Whereas of course you wanted <output.rdf> a :Stylesheet. For that
reason you should probably dispatch on some required part of the
stylesheet language, or there could literally be a magic triple
property.

> Maybe some easy-to-use-batteries-included programming
> language will come for the semantic web

I sorta dread that happening, really, because it's already been
attempted over a span of years with N3Logic which fits exactly what
you've said: it's not a very good programming language by itself. Part
of what I'm doing with the work of which this thread is a spinoff is
trying to engineer a better approach to the generic handling of RDF.

> Anyway, standards are to be enforced or all of us will go crazy
> figuring out every single possibility that might happen.

Yup! Cf. TAG's TagSoupIntegration-54 issue:

http://www.w3.org/2001/tag/issues.html#TagSoupIntegration-54

-- 
Sean B. Palmer, http://inamidst.com/sbp/

Received on Friday, 23 November 2007 10:32:11 UTC