RE: A simple question....

On Thu, 18 Nov 1999, Renato Iannella wrote:
> --On 17/11/99 4:30 PM +0200 Ora.Lassila@nokia.com wrote:
> 
> > Stefan,
> > 
> > you raise a very interesting question. As I recall, it was actually at one
> > time discussed in the RDF working group.
> 
> And I remeber it like it was yesterday ;-)
> 
> I unsuccessfully argued for a facility in RDF syntax to differentiate
> between a URI being used to *identify* another resource (which we have
> now with "resource=" attribute) and a URI that is a link to more RDF
> that _could_ be dereferenced (with a new attribute "metadata=").
> 
> For W3C Memebers you review the thread starting at:
> 
>   http://lists.w3.org/Archives/Member/w3c-rdf-syntax-wg/1998Dec/0007.html
> 
> I still think that this is a _useful_ facility for the "Semantic Web"

See also the 'seeAlso' property in RDF schema, which can be used for
exactly this.	

	2.3.4 seeAlso
	The property rdfs:seeAlso specifies a resource that contains information
	about the subject resource. This property may be specialized using
	rdfs:subPropertyOf to more precisely indicate the nature of the
	information the object resource has about the subject resource. 

This can be used as a hint in the case of self-describing resources.


While syntactic abbreviations to tell us about self-describing uses are
certainly useful, I think it is worth making a point of the fact that
(a) this is just more metadata about something, and (b) there are a
variety of ways of acquiring metadata from a URI-named resource which
we'll want to reflect into RDF. It becomes a slippery slope as to when
'resource=' versus 'metadata=' would've been appropriate.

For example: for many resources in the http:* namespace it might be
appropriate to use the HEAD HTTP method to find out more metadata, or use
WebDAV facilities, or send a content-negotation mimetype preference of
(for eg) text/x-rdf to express an interest in an RDF view of the object.
For resources in the z3950:* URI namespace, your RDF processor might want
to use other mechanisms, eg. the Z39.50 query protocol's EXPLAIN feature,
to acquire more RDF statements about the named resource.

So... hardcoding 'metadata=' into our syntaxes (*not* the
model) might a be useful abbreviation for '[x]--seeAlso-->[x]', or
'[x]--dc:format-->"text/x-rdf"'. But figuring out strategies and commonly
agreeable patterns for acquiring RDF descriptions of a wider diversity of
resources (eg. HEAD, webDAV, EXPLAIN etc) strikes me as a more fruitful
activity.

An example:
If I have an MP3 audio file online somewhere, or a JPEG, or whatever 
data format. Each of these might have embedded metadata, using XML/RDF or
some other older encoding. Using metadata= doesn't help here; instead we
want to figure out how to extract data from each of these formats, either
(?ideally) server side, otherwise, clientside or via some 3rd party
service. Eg. media file URI is http://purl.org/net/danbri/bigpicture.jpeg
We could either download the JPEG, and look inside for metadata. Or we
could pass a reference to it to some specialised RDF Description Service,
eg. 
http://www.tasi.ac.uk/imagemetadataextractor?http://purl.org/net/danbri/bigpicture.jpeg
(which might return us an rdf/xml description of that object). Or else we
could ask the server itself for metadata about that object, perhaps using
content negotiation or WebDAV or HEAD.

Longwinded point being that there's a world of possibilities and the
'metadata=' syntactic suger addresses only a % of the scenarios that
RDF-aware software systems will be grappling with...

Dan

Received on Wednesday, 17 November 1999 22:15:04 UTC