RE: A new proposal (was: Re: which layer for URI processing?)

John Cowan wrote:
>
>
> Jonathan Borden wrote:
>
> >         This is a tricky problem. What is the definition of
> > "R1 equivalent R2"?
>
> I think you mean "How do we decide whether R1 is equivalent to R2"?

	No, I am asking what the precise definition of equivalence is. If you allow
me to define u1 = u2 (literal comparison) and R1 equivalent R2, by
definition, even for relative URIs then problem solved.

> The
> answer comes from outside the system: two URIs identify the same resource
> if the provider of the resource says that they do.

What if the provider of the resource doesn't have a clue (as is literally
true if you care to poll a number of real website providers :-)

"Saying so" may be
> informal, or may be formalized as RDF, thus:
>

>
> >         We have already ascertained that a simple byte
> comparison of R1 and R2
> > won't suffice because the two may have an abstract equivalency. It seems
> > that the equivalency is a priori based on the equality of u1 and u2.
>
> Au contraire.  If URIs are string-equal, they identify the same resource.

	Do we agree on this? If this is true for relative URIs then you agree that
there is no problem with the current Namespace rec, clase closed.

> If URIs are not string-equal, a scheme-specific procedure may exist that
> tells us they identify the same resource anyway.  Beyond that, we
> fall back
> on extracurricular knowledge.  In no case is accessing the resource either
> necessary or sufficient, as I proved in an earlier message.

	You are missing my point. I am arguing to keep the current Namespace rec as
specified. The "problem" which exists is not unique to relative URIs but
also for "absolute" URIs (e.g. "file" "news" "http"?) which are context
dependent.

	Banning relative URIs, or 'absolutizing' them (quotes because prepending
the base URI which itself may be a file URI doesn't cut it) doesn't solve
the problem of context dependency.

>
> >         For example "el cid":

My mistake, scratch that example. Substitute:

file://myschema.xml

This is the same old SYSID vs. FPI issue, revisited.


Jonathan Borden

Received on Friday, 26 May 2000 13:04:51 UTC