- From: Graham Klyne <GK@ninebynine.org>
- Date: Wed, 07 May 2003 10:56:27 +0100
- To: <Patrick.Stickler@nokia.com>, <sandro@w3.org>
- Cc: <uri@w3.org>, <phayes@ai.uwf.edu>
[Nothing of relevance to RFC2396bis herein] At 11:47 07/05/2003 +0300, Patrick.Stickler@nokia.com wrote: >I think that Pat and I actually are in more agreement about this >than disagreement, and in fact the disagreement reflected above >does not exist but arose out of my misuse of the terminology. > >My key point is that, for the SW (and not any application using RDF >that is not considered to participate in the SW) all parties must >agree upon denotations of a given URI which are compatible, even if >not identical. I think one of the continuing terminological problems is this (ab)use of "denotes". In the context of RDF formal semantics, and model theoretic semantics generally (I think), the denotation of a URI is a feature of a particular interpretation, and different interpretations may assign different denotations. This is irrespective of what we may intend a particular URI to mean or represent. As for the intended meaning, I'm still trying to reconcile the ideas that we often think we know what a name is intended to represent, yet, as Pat says, language in general and RDF in particular does not provide any workable mechanism to fix a single meaning for a name. I think there's an interesting parallel between the view of URIs as resource identifiers and formal semantics by model theory. URIs are used to identify some (loosely specified) underlying concept of a resource that yields certain representations under certain circumstances. Those representations are all we actually get to observe -- anything we may wish to know about a resource must be elicited in terms of such representations. Model theoretic semantics likewise does not fix the exact meaning (denotation) of a name, but allows us to constrain its meaning by limiting interpretations to those which match certain statements we may choose to make. What these seem to have in common is a limiting case. The more we say about a resource, the more "closely" it may be constrained to some asymptote of meaning. The more representations of a resource that we examine, the more we may learn about its "essential invariant characteristics". But certaintly is elusive. (Example: let f(x) = sin(x)/x. What is the value of f(0)? We can't evaluate f(0) directly as it involves a division of zero by zero; but if we consider the region about x=0, then a credible argument can be sustained that f(0)=1, for x in radians, because Lim[x->0]f(x)=1.) In the mathematics I've been exposed to, the concept of a limit is strongly related to some idea of a metric, so that we can talk in terms of relative closeness of pairs of values. But there is no obvious metric for denotations of a resource. (At this point I muse about things I don't really understand, such as the theoretical work underpinning aspects of denotational semantics of programming languages ala Scott/Strachey, also ideas of subsumption as appear in description logics...) So can it make sense to think of the intended meaning of a name as a kind of region of meaning, bounded by those things we can observe or say about a resource? Ambiguity remains, but within bounds that we trust will not affect the results we wish to achieve in using a name. There is a kind of presumption here that one can make more observations, assert more constraints, to progressively constrain the nature of a resource in question. #g ------------------- Graham Klyne <GK@NineByNine.org> PGP: 0FAA 69FF C083 000B A2E9 A131 01B9 1C7A DBCA CB5E
Received on Wednesday, 7 May 2003 06:00:14 UTC