Redefining “resource” (was: Re: Drop “g-boxes”, talk about “stateful resources”)

Hi Kingsley,

On 24 May 2012, at 20:43, Kingsley Idehen wrote:
>> I don't think that redefining the meaning of “resource” is realistically achievable at this stage.
>> (Not a comment on whether the term makes sense or not! I just think we are stuck with it. Blame it on the TAG.)
> If a "resource" is anything that can be the referent of a URI, and that definition sticks, we are going to have to live with the resulting confusion and inconvenience.

Okay so I looked around the specs a bit just to see if fiddling with the term “resource” is feasible at all.

I guess the idea would be to call the things denoted by IRIs “entities” or “things” instead of “resources”. And limit “resource” to the REST sense—things that have state that can be expressed in representations, something close in meaning to my proposed “stateful resource”.

We can possibly prod the “resource” via HTTP; we need to apply model theory and social conventions to figure out what the “entity” might be.

So an IRI “identifies” a “resource” and “denotes” an “entity”. These are technically two different relationships with two different mechanisms. Of course, we'd have to say that ideally the two are aligned, and the entity denoted should be the resource identified, otherwise there is ambiguity. Some may decide to reject httpRange-14 and embrace that ambiguity, and might even resolve the ambiguity through some formal punning mechanism that allows a distinction between statements about the entity and statements about the resource.

Could we do this? To see how deep a change this wold require, I looked for uses of “resource” throughout the specs.

  • RDF Concepts doesn't really rely on the term besides defining it.

  • RDF Semantics mostly uses it in the form of the IRI rdfs:Resource.

  • RDF/XML has it all over the place due to the rdf:resource attribute.

  • Turtle almost doesn't use it.

  • SPARQL uses it quite a few times, but nothing too difficult to change.

  • RDF Schema uses it *very* heavily.

  • So does the old Primer.

  • OWL2 barely uses it, except in the RDF-based semantics and in the
    RDF/XML rdf:resource attribute

  • URIs and IRIs are “Uniform/Internationalized Resource Identifiers”, so the
    term implicitly creeps in wherever we talk about URIs and IRIs.

  • RDF is the Resource Description Framework…

I don't know. Sounds like a difficult thing to do and tons of work. Anyone seriously thinks this would be a good idea?


Received on Friday, 25 May 2012 11:28:44 UTC