Re: SemWeb Non-Starter -- Distributed URI Discovery

Stephen Rhoads wrote:
> As far as I can tell, there is no formal, generalized mechanism to
> reliably query the owner of a URI in order to obtain an RDF
> Description of that URI.

While the previous two replies are right that one of the 'features' of 
the semantic web is that there is no authoratative owner for a URI, I 
think there might be some circumstances where a notion of ownership 
would be useful.

Here's a scenario that I've been meaning to seek comment on:

Let's say a blog aggregation service along the lines of something like 
Technorati used dc:subject predicates to categorize blog postings, 
instead of tags.

One blogger writes:
   blogger1:post1 dc:subject <A> .
   Where <A> is the URI to whatever.

And another blogger writes:
   blogger2:post1 dc:subject <B> .

A and B are different URIs, but it's conceivable that they denote the 
same concept and those postings should be categorized together.  How can 
the aggregator know when to consider A the same as B?

Somewhere on the web there has to be an RDF statement:
   <A> owl:sameAs <B> .   (or something similar)

But it can't be that the aggregator will recognize that statement posted 
anywhere on the Internet since that would allow anyone to hijack the 
categorization process.  (I could post 'x sameAs y' for all x and y, and 
now everything gets thrown into one giant category.)

So the question is what to do about a policy decision.  Whose RDF 
statements should the aggregator trust to merge equivalent URIs?

Maybe this is just a mirage, but this seems to divide the set of 
predicates into two groups.  In one group are the predicates that can 
hijack things like categorization, e.g. owl:sameAs.  If the aggregator 
uses OWL-based reasoning, then inverse functional predicates would do 
that too.  That is, if I inject a statement with owl:sameAs or an 
owl:inverseFunctionalProperty, I can affect how the aggregator 
determines what URIs denote the same things, changing categorization. 
OTOH, if I inject a foaf:name predicate, this won't affect how the 
aggregator merges URIs (assuming there's no complex reasoning going on).

The aggregator, if it's going to make use of owl:sameAs to figure this 
out, needs to know who it can trust to make statements that can affect 
the denotations of URIs.

It might be appropriate in this narrow case to consider someone to be 
the owner of a URI, and only the owner of a URI can be trusted to make 
statements about a URI that affect its denotation (relative to the kinds 
of inferences being made by the aggregator).

Let's say that URIs had owners in this sense, and that X owns both URIs 
A and B.  The aggregator comes across the statement A=B.  Its policy 
might be that since both URIs are owned by X, A=B will only be trusted 
if its source is X.  This makes some amount of sense, and might be 
enough sense for blog aggregators in particular to use.

I'm not sure.  But, I think this is something important to think about.

-- 
- Joshua Tauberer

http://taubz.for.net

** Nothing Unreal Exists **

Received on Saturday, 19 March 2005 00:45:14 UTC