RE: interpretations and identifications

Hi Jonathan,

Thanks for the summary.  Responses below.

> From: Jonathan Rees
> 
> I was worried that David and I might have different ideas of 
> what the  
> group's charter or plan was, so I tried to figure out where 
> we agreed  
> or disagreed. The below is obviously my own distorted memory... my  
> apologies to David if I've misrepresented him.
> 
> 1. We agreed that the charter is to publish an ontology of some kind

Yes.

> 
> 2. We agreed that if it got to the point where we were forced to  
> decide between OWL DL and N3 (with rules), that would 
> represent success.

Yes.

> 
> 3. We agree that RDF and/or OWL formal semantics are fundamental.

Yes.

> 
> 3a. We didn't agree whether RDF/OWL were informed or constrained  
> somehow by "web architecture". I maintained that they weren't, that  
> you can conform to RDF semantics without having any respect for "web  
> architecture". Quoth Hayes [1]:
> 
> "The semantics does not assume any particular relationship 
> between the  
> denotation of a URI reference and a document or Web resource 
> which can  
> be retrieved by using that URI reference in an HTTP transfer 
> protocol,  
> or any entity which is considered to be the source of such 
> documents.  
> Such a requirement could be added as a semantic extension, but the  
> formal semantics described here makes no assumptions about any  
> connection between the denotations of URI references and the uses of  
> those URI references in other protocols.
> 
> The semantics treats all RDF names as expressions which denote. The  
> things denoted are called 'resources', following [RFC 2396], but no  
> assumptions are made here about the nature of resources; 
> 'resource' is  
> treated here as synonymous with 'entity', i.e. as a generic term for  
> anything in the universe of discourse."
> 
> I believe it's the intent in both RDF and OWL semantics to cut the  
> cord with the web, and that this is a good thing.

I actually *do* agree with the above.

> 
> 3b. We didn't agree on how RDF might inform web architecture. This  
> hinges on the relation between "denotes" and "identifies" (in the  
> webarch sense).
> 
> Remember that an interpretation is an assignment of things  
> (mathematical, real-world, or otherwise) to URIs. 

Right.  I maintain that RDF semantics is necessary but not sufficient for explaining semantic web architecture.  In particular, RDF semantics intentionally (and correctly, in my view) says nothing about: (a) how a URI becomes associated with the thing (or possible things) that it denotes; and (b) how the range of possible interpretations should be further constrained by additional semantic web architectural principles.  

Questions like "What can be inferred about the resource denoted by http://example/foo if a GET on http://example/foo yields a 200 response?" ultimately make little sense unless there is a coherent way of talking explicitly about the semantics of URIs in semantic web architecture.  The RDF semantics by itself gets you nowhere in answering this question.

Semantic web architecture is layered on web architecture and RDF semantics, and that's what we're dealing with.

> A model 
> theory says  
> what it means for a given interpretation to be a model of an 
> axiomatic  
> system. Web architecture says (in its inimitable way) what it means  
> for a given interpretation to agree with what it calls 
> "identifies" -  
> one might call such an interpretation a model of web 
> architecture or a  
> model of webarch identification.
> 
> There are many legitimate reasons why conformance to constraints -  
> whether logical or not - may be unspecified or underspecified. A set  
> of constraints never specifies a unique model, even in the simplest  
> mathematical situation, or with the most well-crafted prose.
> 
> I would say that "denotes" is a general term to be used for any  
> constraint system, and should only be used either when the system is  
> specified (in context) or aggressively unspecified (as in, say, RDF  
> semantics). I would not want to nail it down to any particular  
> interpretation, given that there may be may be many 
> interpretations of  
> a system under analysis.
> 
> Because it's so useful the way it is, I'd reserve "denotes" for the  
> general metatheory case, not tied to webarch or Euclidean 
> geometry or  
> anything else, and use "identifies" (not a term I like, but 
> following  
> Harry's suggestion to follow webarch) specifically in the theory of  
> web architecture. In a model of webarch, a URI will be said (in its  
> theory, in RDF) to identify what it denotes (in its metatheory,  
> expressed in mathematical prose), but in a model of something 
> else, it  
> won't.
> 
> We have become accustomed to using "denotes" with web 
> architecture, in  
> part because it's superior in many ways to "identify".  But 
> how would  
> you document it then?  You can't talk about the interpretation,  
> because that lives at the metalevel - this would be circular. You  
> could define it to be webarch identification, but then why did you  
> change the label? "Denotes according to an interpretation 
> specified as  
> follows: ..." -- then how do you fill in the ...?
> 
> I'm not seriously proposing that we talk about models of 
> webarch, just  
> introducing the idea to explain myself.
> 
> 4. I proposed we *not* do any metamodeling at this point, 
> meaning that  
> there should be no URI in the ontology for "denotes", 
> "models", etc.,  
> as this risks being circular, even if we can figure out which  
> interpretation is meant. David disagreed, saying that otherwise we  
> don't know what RDF we're working with, or what URIs in 
> general mean.  

Right.  We cannot talk sensibly about "deprecating a URI", and make statements that relate the meaning of the old URI with the meaning of the new URI unless we can talk explicitly about the association between a URI and what it denotes.  For example, if http://jann.example#akt was minted to denote a newly discovered protein, AKT, which is later determined to be three distinct proteins, then that URI may be deprecated in favor of three new, more specific URIs: http://luke.example#akt1 http://luke.example#akt2 and http://luke.example#akt3 .  In which case, it would be helpful to indicate in RDF that the original URI has a broader range of interpretations than the new URIs:

"http://jann.example#akt"^^xsd:anyURI s:isBroaderThan 
    "http://luke.example#akt1"^^xsd:anyURI ,
    "http://luke.example#akt2"^^xsd:anyURI ,
    "http://luke.example#akt3"^^xsd:anyURI  .

This is described in
http://dbooth.org/2007/splitting/

In short, I see no way around talking about what it means, in semantic web architecture, for a URI to *denote* a resource.

> I said let's just assume for now that have a ready supply of it, and  
> work instead on figuring out what it would mean, if we had 
> it. There's  
> plenty to do before we get blocked for want of a formal 
> metatheory of  
> axiom set synthesis and denotation.
> 
> (OK, now to hope that Pat doesn't give his pupil's essay a bad grade.)
> 
> Jonathan
> 
> [1] http://www.w3.org/TR/2004/REC-rdf-mt-20040210/
> 



David Booth, Ph.D.
HP Software
+1 617 629 8881 office  |  dbooth@hp.com
http://www.hp.com/go/software

Statements made herein represent the views of the author and do not necessarily represent the official views of HP unless explicitly so stated.

Received on Wednesday, 4 February 2009 18:17:33 UTC