W3C home > Mailing lists > Public > www-tag@w3.org > September 2007

Re: ISSUE-58: Scalability of URI Access to Resources

From: Pat Hayes <phayes@ihmc.us>
Date: Fri, 07 Sep 2007 12:32:09 -0500
Message-ID: <46E18B19.8080309@ihmc.us>
To: Chimezie Ogbuji <chimezie@gmail.com>, www-tag@w3.org, skw@hp.com

Chimezie Ogbuji wrote:
> On 9/7/07, Williams, Stuart (HP Labs, Bristol) <skw@hp.com> wrote:
>   
>> Hello Chimezie,
>>
>> So... I'm going to be a little obtuse, maybe deliberately so... but
>> maybe I should also make it clear what is bugging me, which is the use
>> of the prefix RDF in the term "RDF URI"... they are just URIs... I don't
>> see what insisting on calling them "RDF URI" buy's you.
>>     
>
> I see I have failed (perhaps miserably) in making this point.
I also don't really understand your point here, Chimezie, and agree with 
Stuart that URIs in RDF are, well, simply URIs. After all, the same URI 
can be used in some RDF and also in another context altogether, say in 
HTML. Its true that RDF(/OWL/etc.) places different semantic constraints 
on URIs than non-SWeb applications do, but thats not because the URIs 
are different.
>> I'm afraid that didn't help me :-(. I certainly understood the semiotic
>> triangle stuff and have been exposed to it before.
>>
>> As 'symbol' what makes "RDF URI" (your term)  different from URI in
>> general.
>>     
>
> See below.
>
>   
>>> This is mostly a rehash from http://www.w3.org/TR/rdf-mt/ (so
>>> I hope Pat will slap me on the hand when I'm chatting
>>> rubbish), but generally, RDF URIs are 'symbols' which denote
>>> things in an interpretation (a 'theory').
>>>       
>> And URI in general do not? Even in the particular interpretation that we
>> might commonly call 'the web'?
>>     
Stuart, the Web isn't a particular interpretation. Its a rather large 
'theory' in the logical sense which needs to be interpreted. And (maybe 
this is what Chimezie means?) if the Sweb hadn't come along, I don't 
think anything in the entire Web world would have suggested that URIs 
/denote/ anything at all. What they do is connect you to resources, not 
denote them.
>
> URI's in general denote 'resources' but (formal) denotation IMHO is
> only as useful as the inference capabilities it facilitates.  If this
> wasn't the case, there would be no reason to bring KR into the picture
> (i.e., 'the web' + KR = 'semantic web')
>
>   
>>> The expressions in
>>> which the RDF URIs are used describe a set of conditions that
>>> must be met to satisfy the interpretation.  The end-game
>>> (goal, if you wish) is "to provide a technical way to
>>> determine when inference processes are valid, i.e.
>>> when they preserve truth."
>>>       
>> I don't understand how I can tell whether or not I'm using an "RDF URI".
>>     
>
> >From the context.  An unparsed RDF/XML document is nothing more than
> an information set, with QNames composed from URIs.  The namespace
> names denote 'resources' (per URI RFC), but this denotation is
> basically inert: there is no mechanism to guide inference (or
> interpretation).
>   
Put another way, you have (sound of teeth gritting) denotation but no 
assertions. You only get inferences made valid when some sentence 
actually says something ABOUT the denotation. If all you have is the 
denoting name, then about all you can do is to repeat it more or less 
loudly.
> After you parse the document into an instance of the RDF abstract
> model, you have a different context and different mechanisms apply (by
> normative specifications).  They are still URIs in the RFC 2396 sense,
> but the denotation is no longer inert.
>   
I hope/think I see what you mean, but this is a very odd way of putting 
it. How about: what the RDF gives you is some actual formal assertions 
about the referent, from which you might be able to draw some new 
conclusions about it.
>   
>> OTOH, I can tell when I'm using a URI in RDF (or HTML or an email or...)
>>
>>     
>>> The main difference is that the domain of discourse is a
>>> superset of what Web architecture is primarily concerned
>>> with: information resources and their representations.
>>>       
>> Though I think Pat would wish it otherwise, I don't think that in terms
>> of the denotation of URIs, the domain of discourse in Web Arch is
>> constrained in the way that you describe.
>>     
I think it is, in fact. Without the Sweb, all claims about URIs denoting 
non-information resources is very little more than a TAG fairy story, 
and about as easy to demolish as any other fantasy.
>
> Note, I said 'primarily'.  Consider that a majority of the 'Good
> practice' assertions in AWWW are relevant only to representations.
>
>   
>>> Whereas in a model-theoretic language, the 'semantics' are
>>> determined from the expressions which make use of the RDF
>>> URIs, Web architecture is (or it seems that way from the
>>> specific best practices in AWWW) primary concerned with the
>>> consumption of information resources to meet a different
>>> goal: a user browsing a page, or a web crawler browsing pages
>>> to create indices for subsequent searching.
>>>       
>> Hmmm.... I'll think about that. There is a particular artiefact, AWWW, a
>> document which is an expression of a set of principles and best
>> practices that we could agree on (at the time). However, there is the
>> much larger conceptual artifact of "The Architecture of the Web". Whilst
>> AWWW speaks mostly of what it calls "information resources" the scope of
>> what URIs can (and I'll use the word here risking a blast from Pat)
>> identify (by which I mean 'refer-to') is unconstrained by AWWW.
>>     
It is unconstrained, but there is no reason to suppose that it is 
possible. Take the entire Web pre-RDF, and tell me where there is 
ANYTHING in its architecture or usage which requires a URI to 
denote/identify/have any meaningful relation to anything other than an 
information resource. The fact that some Web pioneers had wet dreams 
about revolutionizing the Nature of Meaning doesn't tell us anything 
about anything other than psychology and ambition.

Pat
>
> Yes, they are certainly unconstrained, since it essentially describes
> a 'resource' as something which is consistent with arbitrary
> referents.  It is not so much the constraint (or lack thereof) that
> bothers me but when attempts are made to make authoritative claims (or
> suggest that such claims can be 'found' - hence the load on the
> transport level and the original point of relevancy for ISSUE-58 )
> about the nature of referents independent of a formal inference
> mechanism.
>
>   
>>> By the context of its use.
>>> http://metacognition.info/profile/webwho.xrdf#chime is the
>>> URI I've 'minted' to represent me.
>>>       
>> ok...
>>
>>     
>>> When used as a link in an
>>> HTML document, it is simply a (typed) link to my FOAF
>>> document.
>>>       
>> Hmmm... less ok... assuming that you are serving as
>> "application/rdf+xml" the by virtue of the RDF media type definition
>> that is still a reference to you, the person, rather than a bit of text
>> in an RDF/XML representation of a graph - or more accurately it is a
>> reference to a/the thing in an intepretation which statisfies the
>> assertions made in the graph (which in all probability is you).
>>     
>
> Point well taken, my example was poorly-formed and the media-type
> gives you enough information to essentially 'blur' the distinction.
>
>   
>> FWIW: I believe that "unique denotation of URI" is an intention the
>> design of Web Architecture; I am equally aware that such a position
>> denies "context of use" as a means to disambiguate the referrents of
>> referring expressions. I am equally aware that in general the context in
>> which a referring expression occurs can be significant. I have never
>> been able to square that circle - though the closest I have come is to
>> regard the web as one big context and referrents of references made
>> using URI as invariant within that (large) context.
>>     
>
> That is a very difficult circle to square, yes.  However, abstracting
> the web as one big context works against interpretation - since
> something more sophisticated (i.e., KR) is needed for this.  I would
> suggest that if the context is significant *precisely* because it
> gives a (mostly) unambiguous path to determining the nature of
> referents  then its usage should not be denied.
>
>   
>> It seems like a
>> design choice in the design of Web Architecture - it is I think the
>> place where the constraints of the Web apply to the semantic web.
>>     
>
> Hmm.. /me will need to sleep on that one =)
>
> -- Chimezie
>
>
>
>   
Received on Friday, 7 September 2007 17:32:23 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:56:17 UTC