Re: My action item on RDDL/RDF

Dave Beckett wrote:
> ...
> I normally just lurk here but I couldn't ignore this flame-bait.

How is it flame-bait? I thought it was accepted by both the RDF and XML 
worlds that RDF/XML is brutal. Presumably the point of RDF/XML was to 
make RDF acceptable to XML people. It has been a huge failure. I don't 
see your name on the original spec so I don't know why you take it 

> I edit the RDF/XML specification (revising the existing syntax) so
> here is how I see it.  One of RDF/XML's many original goals was to
> enable it to be embedded in HTML (Tim Bray was on the original
> working group and can correct me).  This is one of the reasons why it
> has many (too many) abbreviated forms that allow you to express
> statements without using element content; dumb HTML browsers will
> ignore it.

I don't see anything in the grammar that allows you to put string values 
in attribute values rather than content so it is a huge failure if 
legacy browser support is a goal.

[6] propertyElt    ::= '<' propName '>' value '</' propName '>'
                        | '<' propName resourceAttr '/>'

[8] value          ::= description | string
[9] resourceAttr   ::= 'resource="' URI-reference '"'

So how do I hide a string from an XHTML parser?

A *HUGE* problem is that RDF/XML seems not to have any easy way of 
having either literal XML or mixed-content as value (to say nothing of 
other type-declared data!)! Maybe if this was fixed I could get used to 
RDF/XML. Jonathan had a RDDL-specific fix for it but this really needs 
to be fixed in RDF itself. Without that, RDF's XML syntax is not just 
inconvenient, it is actually hostile to normal ways of using XML! And 
surely any rational integration of XHTML and RDF has to address XHTML 
inside RDF in addition to the reverse!

As an aside, it would be a lot easier to follow the striping if the 
specs suggested naming (e.g. upper/lower-case) conventions. I'm tempted 
to go as far as the old CToolBar convention. XML syntax doesn't 
distinguish between objects and properties so it is very easy to get lost.

Another issue: I find it odd that there is no "unabbreviated" syntax for 
properties in RDF/XML. RDF/XML would conceptually simpler if it started 
with a very simple, obvious triple syntax as Tim Bray suggests and then 
adds abbreviations rather than starting out half-abbreviated.

> Yes, and that is a good approach.
> For the record:
>    I do not expect all RDF graphs to be written in RDF/XML.
> and indeed, I'm planning to add mapping code to my RDF parser by the
> means of the XML mapping or transformation language, XSLT.  This will
> allow scraping from any XML tag soup such as the RSS* ones ;)

XSLT is a very heavy stick. It is a full Turing complete language, it is 
not really discoverable from XML instances, it isn't streaming-friendly. 
I don't think that the semantic web should depend on XSLT for mapping 
XML to RDF. IMHO, there should be a first class XML->RDF mapping language.

  Paul Prescod

Received on Monday, 11 November 2002 20:50:38 UTC