Re: Site metadata; my preference

On Wed, Feb 19, 2003 at 04:50:59PM +0100, Yves Lafon wrote:
> > Exactly.  "meta-html" is still HTML, so should be using the text/html
> > media type.  Conneg is for handling variability in representations,
> > not variability in resources.  The latter is what URIs are for.
> 
> I use Accept: application/rdf+xml for now, even it we can't restrict
> semantic to a single mime type. (I don't want to open the issue about what
> mime type is _really_ ok for a resource, also ;) )

8-)

But "Accept: application/rdf+xml" says "Send me an RDF representation
of this resource", which is different than "Send me an RDF
representation of the metadata about this resource".  You wouldn't
expect that if "Accept: text/plain" gave you "I'm Dan's car", and
"Accept: image/gif" gave you a picture of the car, that
"Accept: application/rdf+xml" would give you some RDF about the
metadata about the car; you'd expect RDF about *the car* (or
more technically, the state of the car), such as its colour,
its location, its mileage, etc..

RDF gets a bum rap as a "metadata" solution, when it's a perfectly good
data solution.

> > To Yves;
> >
> > Re OPTIONS, that's a good example, but it appears to me (as I've used
> > it quite extensively), that it falls on the other side of the equation
> > that evaluates the trade offs with respect to latency.  In the uses I
> > made of it, an extra round trip was a non-starter.
> 
> Well OPTIONS is intended to be extensible, may have a body both in request
> and reply (even if it is undefined right now) and is mainly about
> cummunication options on either the esrver or a specific resource.
> But what is communication option?

Some aspect of the protocol in use.

> In the light of a HTTP URI, which is to
> me an HTTP view of a more generic object, it is metadata of this
> particular HTTP view of the resource. Also when you get a representation
> of the metadata of a resource you only get a facet of it. How the metadata
> of the HTTP view and the HTTP view of the metadata of this object collide?

They collide by not being representations of the same resource.  For
one, it makes assertions with the URI ambiguous; are they about the
data or the metadata?  (as has been discussed ad nauseum)

> If they are _clearly_ distinct, then OPTIONS has its use as being part of
> a subset specific only to HTTP and accessible only via HTTP means.

OPTIONS isn't about resource metadata, it's about communication options,
like which methods are implemented.  It could be extended to communicate
WSDL-like stuff, like the acceptable/supported data formats, or other
supported protocols, protocol upgradeability, etc..  None of those
have anything to do with resource metadata, AFAICT.

MB
-- 
Mark Baker.   Ottawa, Ontario, CANADA.        http://www.markbaker.ca
Web architecture consulting, technical reports, evaluation & analysis

Received on Wednesday, 19 February 2003 11:03:49 UTC