RE: working around the identity crisis

I think there are four key issues floating around here:

1. Using distinct URIs to identify distinct resources (and I think
most everyone is agreed that that is a necessary thing to do).

2. Having a shared set of representations between multiple resources
such that that same set of representations is accessible via multiple
URIs, by some means.

3. The means by which distinct resources are related, such that they
share the same set of representations; e.g. resolution of a concept URI
resulting in redirection to a document URI, either by conneg or some
other means (see below).

4. Autogeneration/automapping of one form of representation to another
(e.g. generating HTML from RDF/XML, etc.)

It's important to note, I think, that only the first issue is
"architectural" in nature. The rest concern "localized" methodologies
employed by particular systems, and as such, are (or should be)
opaque to the web architecture proper.

I.e., while particular publishers will employ consistent methodologies
to organize content/knowledge and process requests, and while those
methodologies will likely introduce human-discernable characteristics
into the lexical form of the URIs employed, web clients (and humans)
should be repeated warned against making any presumptions about the
identity or nature of resources based on any of (a) the lexical form of
URIs, (b) server behavior when resolving URIs to representations, or 
(c) the nature of the representations returned.

That said, I thought I'd offer an example of an alternative approach
to using conneg and 'suffixes', which is based on metadata-defined
relationships and which is independent of the lexical properties of
URIs (this is the approach we use for many resources accessible
via the Nokia Semantic Web Server).


The URI identifies a concept "Foo". 

The URI identifies a document
which provides information about the concept.

Both share the same representation (i.e. resolving either of the
above URIs will result in the same octet stream).

However, the relationship between those two resources, that 
the first shares the representations of the latter, are defined
in terms of the property
rather than conneg.

When direct resolution fails, the SWS looks to see if any such
relationship is defined, and if so, redirects to the related

In essence, it is a form of implementation of the PURL concept,
but independent of the notion of 'download location' or the like, 
but is defined generically in terms of the relationships between
resources and their web accessible representations.

And of course, there's no reason why this resolvesAs approach
cannot be used in conjunction with conneg and other methods,
depending on the particular needs of the system and what offers
the greatest utility for specific cases.


There are also a few interesting things to note about

* It is an information resource (see the metadata description)

* Its entire substance is included in the representation returned
  (as the creator of both, I assert that this is true)

* There is additional information included in the representation
  returned which is not part of the substance of the information
  resource (try to guess which bits are part of the information
  resource and which bits are 'extra', I bet you can't ;-)

Anyway, just thought I'd offer that alternative perspective...



> -----Original Message-----
> From:
> []On Behalf Of ext Miles, AJ
> (Alistair)
> Sent: 14 November, 2004 19:20
> To: ''
> Subject: working around the identity crisis
> Sorry, reposting this with a more sensible title - a discussion about
> working around the 'identity crisis' for http uris:
> Would very much like to know what folks think of this.  Have 
> I just managed
> to redescribe some emerging consensus, have I misunderstood 
> anything, should
> I go away and read more?
> I wanted to say also that I am acutely aware that alot of 
> people have spent
> a lot more time thinking and writing about this than I have.  
> I'm really
> just trying to understand this problem space, and get to a 
> place were I can
> effectively explain the issue (and the options) to other people.
> Responding to Patrick:
> 'If you have a URI that identifies a concept, and 
> dereferencing that URI in
> a browser results in some web page displayed in that browser, 
> that does not
> mean that the URI has been used to identify two things, the 
> concept and
> the web page (document).'
> I think that puts very concisely the fundamental point I was 
> trying to make.
> I probably didn't need to say any more than this.  
> Responding to Daniel:
> 'A good solution, but I still prefer the notion of using XSLT to
> transform the RDF/XML definition of a concept into a human friendly
> HTML page about it.'
> I agree that, as a matter of good practise, any alternate content-type
> representations of the same concept need to be synchronised.  
> One way of
> achieving this would be to use an RDF/XML description as the reference
> point, and using XSLT to generate an HTML representation.   
> Yours,
> Alistair.
> ---
> Alistair Miles
> Research Associate
> CCLRC - Rutherford Appleton Laboratory
> Building R1 Room 1.60
> Fermi Avenue
> Chilton
> Didcot
> Oxfordshire OX11 0QX
> United Kingdom
> Email:
> Tel: +44 (0)1235 445440

Received on Monday, 15 November 2004 09:40:48 UTC