RE: CURIEs: A proposal

>>In effect, you are saying that usage determines meaning, not 
>>architectural principles.  A URI in a SemWeb doc is treated
differently 
>>than when in other situations.

>Yes, exactly. Though not *all* URIs in SWeb docs, eg owl:imports uses
GET and so do rdfs:seeAlso and a few others. These >uses are
conventional Web technology. (To properly align these uses with logical
semantics takes some extra work in the 
>specs which is in fact missing from the RDF and OWL specs, but is
included in the RDF 'named graph' proposal and the more >recent Common
Logic ISO draft.)

This is almost an architectural form approach (see HyTime) in its
effect.  Another way to look at it is locale and subject view determines
semantic (Peircean).  Too exotic for most tastes because it insists on
the web being a system of systems with the upper-level ontology being
very light much as Sowa insists it has to be.

>>  Local system prevails.

>Yes, although what is in the 'local system' might vary from use to use.


Which is what makes me think of Peirce with a timestamp.

>The meaning of logical names is a function of the totality of
descriptive capacity that is brought to bear on the task on >making
inferences, and this might involve RDF/OWL content brought together from
any number of sources; so 'local' should >not be thought of as
document-centric: local to the current context of use of the name.

The combined subjective views determines the objective space.  Logical
but dynamic and not nearly as orderly or predictable as some want from a
so-called semantic web particulary if there is feedback from the
objective space into the subjective views so they begin to behave like
infinite impulse response filters (see signal theory).

>Ok but it means there is an operational disconnect between two 
>applications of the same abstraction: the URI.

>Right. The architectural use *requires* the Web infrastructure; the
inferential use only requires that the names have 
>global scope.

In practice, each use can have its own architecture.  Interoperability
is lattice-like.  Again, Sowa would approve.

>>It also means the advice from some to use URNs instead of HTTP URIs
for 
>>namespaces is good advice.

>Well, it is kind of irrelevant from a strict inferential perspective;
but I take your point.

It is pragmatic.  The distribution of the URIAsActiveControls
(clickable) is far greater than the distribution of URI as
NamesNotDereferencedInPractice because the use of URN catalogs isn't
that ubiquitous.  So if use determines semantic and that is a density
function, URN for Non-Dereferenced Name is a pretty good bet.
Hysteresis acts as a boundary builder.

>I should say for the record that the line I have been arguing here is a
very strict interpretation of the logical role of
> URIs. Not everyone in the SWeb world would agree with me, and
sometimes I don't agree with myself. I wish the SWeb were
> better integrated with the conventional Web, and the two uses of URIs
were more closely integrated (I see RDF-A as a
> step in this direction). But we will not achieve this richer future
state merely by wishing that it were true: 
>we need to put mechanisms in place to support it.

Hmm.  Mechanisms or best practice?  The weaker form of the architectural
finding is that anytime one uses a URI that is likely to be viewed by a
human in a browser, one should put a document at the location it
identifies when dereferenced not because the system demands it but
because it is polite to the human given the tool used.  In short, a
SemWeb tool won't turn it blue and make it clickable but the act of
copying it into a web browser address field will.   So there is no
getting around the context of volitional use.  The URI obtains
operational affordance from the norm space it is in at some location in
time. 

It is the trap of insisting that all names are identifiers where
identification is the act of resolving a reference to some information
resource that must exist on the web where the web is an abstraction for
the universal set.

len

Received on Thursday, 29 June 2006 20:56:43 UTC