- From: David Booth <david@dbooth.org>
- Date: Tue, 06 Apr 2010 23:03:26 -0400
- To: Jonathan Rees <jar@creativecommons.org>
- Cc: AWWSW TF <public-awwsw@w3.org>
Explanation of my answers to the doodle poll: Regarding option #1, I disagree with option #1, because I think it is fine and extremely useful to use URIs as names for any kind of things. By issuing 200 responses, Alice and Bob both are indicating that their URIs denote information resources. But is fine for a URI to name something that has characteristics both of being an information resource and something else, because there is always ambiguity in determining what a URI denotes. This is further explained in http://dbooth.org/2007/splitting/#httpRange-14 Regarding option #2, I cannot determine whether Bob is doing anything wrong, because the scenario does not indicate whether Bob is actually serving anything that is in *conflict* with what he has said about the FRBR expression his URI denotes. If the assertions that he makes via HTTP 200 responses are in conflict with what he said the URI denotes -- or indeed, what Alice said the URI denotes, since he has also said that "<http://example.org/a> owl:sameAs <http://example.org/b>" -- then he is doing something wrong: he is contradicting himself. Otherwise what he is doing is perfectly fine. Regarding #3, I'm not entirely sure I understand #3, but if #3 is saying that HTTP responses are completely independent of what the URIs denote then I disagree with #3. Regarding #4, I completely disagree with #4, since such a huge part of the value of URIs as names is their context independent nature. ---- A key question is: how does a URI (as a name) become bound to a resource? And for whom? The AWWW grants a URI owner the authority to specify the binding between that URI and a resource: http://www.w3.org/TR/webarch/#uri-ownership But really all this means is: (a) one should not squat in someone else's URI space, by trying to say what that other person's URI binding should be; and (b) if you want to know the intended binding for a URI, you should listen to its owner. The owner has many ways of saying what that binding is intended to be. One way is by issuing HTTP responses to GET requests on the URI. Another is by making out-of-band pronouncements, as Alice and Bob did in a coffee shop. It is a fallacy to think that a URI denotes only a single resource. Yes, by design a URI is *intended* to denote one resource: http://www.w3.org/TR/webarch/#id-resources And that level of understanding is good enough for most people and most cases. But if you dig deeply into the semantics of web architecture (which you are doing) I think it is much more helpful to think of URI denotation in terms of RDF semantics, which makes it quite clear that a URI does *not* necessarily denote a single resource. Rather, a URI denotes a single resource only in a particular RDF *interpretation*: http://www.w3.org/TR/rdf-mt/#interp But a given RDF graph URI may have *many* interpretations. Hence, there may be many resources that correspond to the same URI. The important thing is that the set of permissible interpretations *constrains* the set of resources that the URI may be interpreted to denote: http://dbooth.org/2009/denotation/ This is *not* the same thing as saying that a URI is context sensitive. Rather, it is merely acknowledging the ambiguity of reference that is inescapable when one attempts to define or decide what resource a URI denotes. We do not have any practical way to directly bind a URI to a *single* resource. But what we *can* do is to indicate a set of assertions that suitably *constrain* the permissible interpretations associated with a URI. This is the purpose of a URI declaration: http://dbooth.org/2007/uri-decl/ Whether or not one agrees with the best practice of serving the URI declaration through the URI's "follow your nose" location http://dbooth.org/2007/uri-decl/#nose I do not know of any other practical way to specify the intended binding between a URI and a resource, other than by somehow providing what amounts to a set of assertions that constrain the URI's resource identity. One cannot point at everything one wishes to denote, saying "this URI denotes *that* person" or "this URI denotes *that* protein". David Booth On Wed, 2010-03-31 at 10:52 -0400, Jonathan Rees wrote: > I've set up a Doodle poll, named(??) by the URI obtained by > concatenating the following two strings: > > "http://doodle.com/qcygav3k8ctmht" > and > "z4" > > (yes, I know this list is publicly archived, I'll accept the risk). > The poll will close in about a week or when it starts getting spammed, > whichever comes first. > > I'd really like for everyone reading this list (all 14 of you) to > weigh in. It's my best way to know I'm not talking into the wind, and > get an idea what the received wisdom is supposed to be. > > Jonathan > > > -- David Booth, Ph.D. Cleveland Clinic (contractor) Opinions expressed herein are those of the author and do not necessarily reflect those of Cleveland Clinic.
Received on Wednesday, 7 April 2010 03:03:55 UTC