RE: Resources and URIs

Simple. One may wish to use the existing global infrastructure of
Web servers to serve other than representations of resources, such
as metadata descriptions of resources.

If you use some other URI scheme than http:, you will also have
to design and deploy a solution for the global resolution of those
URIs.

Think of it this way. An HTTP server resolves requests based on
URIs to responses of various kinds. The mode of resolution applied
to that URI might involve representations (the present default
mode of resolution) or it might be some other mode, such as
involving metadata descriptions of the resource (such as for
a SW agent).

In either case, one may use the existing global HTTP infrastructure
and simply extend ones HTTP server to recognize and support new
modes of resolving URIs.

Note, though, that different modes of resolution do not mean that
the URI denotes different things. Rather, it simply means that
you are asking different things of the server with regards to
a particular resource, which is denoted by the URI.

So, if we have URI-Resolution-Mode: Representation, then we are
asking the HTTP server to give us a representation of the resource
denoted by the URI, a snapshot in time of the state of the resource.
If we have URI-Resolution-Mode: Description, then we are asking the
HTTP server to give us a metadata description of the resource (the
encoding of which can be specified with standard conneg machinery,
but defaults to RDF/XML). In both cases, the URI denotes the same
resource. We are just interacting with that resource via the server
in different ways.

Thus, given all the billions (trillions?) of existing resources
denoted by http: URIs, the servers hosting those resources can
easily begin to also provide metadata descriptions of those resources
in a standardized, globally scalable manner simply by being extended
to recognize and support the new mode of URI resolution.

Cheers,

Patrick


PS: credit to the use of a single HTTP header to indicate the
mode of resolution goes to Sandro Hawke, with many thanks ;-)


> -----Original Message-----
> From: ext dan@tobias.name [mailto:dan@tobias.name]
> Sent: 29 April, 2003 17:17
> To: uri@w3.org
> Cc: Stickler Patrick (NMP/Tampere)
> Subject: RE: Resources and URIs
> 
> 
> On 29 Apr 2003 at 15:30, Patrick.Stickler@nokia.com wrote:
> 
> > > My understanding of his view is that http: URIs can only name 
> > > documents, unless they're URIrefs containing # in which case they
> > > can name things of any kind.
> > 
> > Well, I'm also familiar with Tim's recent arguments along these
> > lines, though I think he is in this case in a pretty small minority.
> > 
> > If URIs can denote anything, then one need not use a URIref to 
> > denote something that is not a "web document".
> 
> I don't really understand why anybody ought to use "http:" URIs to 
> denote anything other than the document dereferenceable via that URI 
> and the HTTP protocol.  That seems to be asking for confusion.  Why 
> not use a different URI scheme if one needs URIs to denote non-
> dereferenceable entities such as persons or organizations?
> 
> -- 
> Dan
> Dan's Web Tips: http://webtips.dan.info/
> Dan's Domain Site: http://domains.dan.info/
> 
> 
>

Received on Friday, 2 May 2003 03:20:46 UTC