Re: Clarifying what a URL identifies (Four Uses of a URL)

Michael Mealling wrote,
> > On the one hand you're saying that a URI identifies a Resource
> > across contexts.
>
> No. I'm saying that URIs and Resource exist regardless of contexts.

But presumably you agree that a given URI identifies a particular 
Resource? If not then why the "I" in "URI"? Or are you saying that a 
given URI might identify different Resources in different contexts/at 
different times or whatever?

If the latter, then I think you're closer to Sandro's position than you 
think.

> > On the other you're saying that a URI is always used in a
> > given context, and that contexts can vary, so the effects of use
> > might vary correspondingly.
>
> It depends on how sloppy your context is with its URIs. If a context
> says that URIs can be anything and you go redefining and adding stuff
> that's inconsistent then, yea, your effects will varry
> correspondingly.

Well, take one very specific example. An XML namespace identifier 
identifies an abstract namespace, and, if any of the various namespace 
document proposals make any headway it'll also identify the document.

That might be sloppy, but it seems to be being wired into the Web 
Architecture.

> > That's not flat out inconsistent ... tho' it would be if we agree
> > that meaning (ie. the referent) is determined by use ... but it
> > certainly looks like a very unstable position.
>
> Its not. What you're doing is trying to assume some part of the
> infrastructure has more meaning than it really has and when that
> meaning your trying to attach is inconsistent you then blame the
> infrastructure compoenent instead of the sloppy application of
> meaning.

I'm not sure what Sandro's position is, but I don't think _bare_ URIs 
have any meaning at all. But I do think that URIs are meaningful in the 
context of their use. For example, in namespace processing contexts 
http URIs denote abstract namespaces. And in HTTP retrieval contexts 
they denote a communication endpoint. In other contexts they might 
denote bunches of bits, or documents, or printers, or organizations, or 
concepts, or whatever.

There's also often an implicit context (eg. if I mention 
http://www.w3.org/ in a mail to a mailing list there's usually a 
pragmatic implication that the HTTP retieval context is in play). But 
implicit context tends to be in the eye of the beholder, which renders 
URIs which are only implicitly bound pretty much ineliminably ambiguous 
... tho' mostly harmlessly so.

I think Sandro's problem is that the use of URIs in RDF is to a large 
extent infected by this kind of ambiguity, and here the ambiguity is 
harmful. And I think he wants to have the Web Architecture recognize 
that ambiguity, partly just to tell it how it is, but also partly to 
underwrite the addition of disambiguating mechanisms to RDF.

I think he's right on all counts.

Cheers,


Miles

Received on Tuesday, 21 January 2003 17:16:00 UTC