W3C home > Mailing lists > Public > uri@w3.org > July 2003

RE: Proposal: new top level domain '.urn' alleviates allneedforurn: URIs

From: <Patrick.Stickler@nokia.com>
Date: Wed, 16 Jul 2003 12:15:10 +0300
Message-ID: <A03E60B17132A84F9B4BB5EEDE57957B5FBC05@trebe006.europe.nokia.com>
To: <michael@neonym.net>
Cc: <sandro@w3.org>, <hardie@qualcomm.com>, <uri@w3.org>

> -----Original Message-----
> From: ext Michael Mealling [mailto:michael@neonym.net]
> Sent: 10 July, 2003 18:56
> To: Stickler Patrick (NMP/Tampere)
> Cc: sandro@w3.org; hardie@qualcomm.com; uri@w3.org
> Subject: RE: Proposal: new top level domain '.urn' alleviates
> allneedforurn: URIs
> On Thu, 2003-07-10 at 11:14, Patrick.Stickler@nokia.com wrote:
> > > Its all about _layers_ Patrick. 
> > 
> > No shit. Don't patronize me.
> Sorry 'bout that. I'm just seeing layer violations in just about
> everything you're suggesting.....

Then you must be misunderstanding me, or I must have no clue
as to what a "layer violation" is for you (or maybe a bit
of both).

> > You explicitly stated that two lexically distinct URIs denote
> > *different* resources. 
> Given the lack of language in 2396 about comparing resources 
> I can't see
> how you can suggest otherwise.

Errr... if 2396 does not talk about comparing resources, that that
also means that you cannot presume that two lexically distinct
URIs *definitely* denote different resources -- only that the two 
URIs *might* denote different resources.

I also see no reason why, given knowledge of the significance
of 2396 to the Web and SW architectural layers, that these 
points cannot be made clear, even without referring to those
higher layers.

I.e. RFC 2396 can explicitly point out that 

(a) there is no requirement that there be a 1:1 correlation between 
resources (things identified by URIs) and URIs -- i.e. it is
possible that lexically distinct URIs may be used to identify
the same thing.

(b) the only equivalence test RFC 2396 provides is for lexical 
equality of URIs, and that is not sufficient to determine whether
two lexically distinct URIs denote the same resource, only
whether two URIs are lexically the same.

Saying the above would provide a clearer foundation upon
which higher layers such as HTTP, RDF, OWL, etc. provide
the extended machinery to clarify the meaning of and
relations between different lexically distinct URIs.

HTTP can define equivalence of denotation between URIs 
which differ lexically in particular ways, such as in
case for the scheme and web authority prefix, etc.

OWL can define equivalence of denotation between arbitrary
URIs entirely.

Yet despite those extensions which allow one to equate the
denotation of lexically distinct URIs, the definition of
what a resource is (anything in the universe) and the
relation between a URI and a resource (the URI refers to
or denotes the resource) can remain the same for all layers.

> > That's simply wrong. And means that the lower layer is
> > constraining higher layers against co-denotation.
> No. I'm simply saying that higher layers have to create new 
> definitions
> for 'resources' that are a superset of the one used by 2396. 

But if the definition of 'resource' could be identitical for
all layers (and I assert that it must) wouldn't that be better?

> I.e. there
> are URI-Resources and there are SW-resources (or LDAP-resources, or
> Gopher-resources). A SW-resource can be referred to as a URI-Resource
> but if that is done by something that isn't SW aware then it 
> must not be
> expected to 'know' about SW concepts of equivalence.

You seem to be equating resource and representation again. That
a URI resource is a particular bit-stream.

I see no reason to define 'resource' in such a constrained manner,
nor does such a definition seem compatible with any of the higher

I also fail to see how the present definition of 'resource' in
RFC 2396 fails to correspond to an LDAP-resource, or Gopher-resource, 

> > It is true that RFC 2396 is unnable to say anything about 
> equivalence
> > of denotation. Fine. But it should also clearly reflect that
> > there is no restriction against co-denotation of URIs.
> Correct. There is no restriction of co-denotation because 2396 is
> completely silent on subjects like that. 

Well, I'm glad that 2396 will remain agnostic about co-denotation,
but I was taking issue with your assertion that lexical inequality
of URIs could/should be taken to indicate distinct denotation.

Perhaps you didn't mean to make that assertion. Perhaps there really
is no point of disagreement.


> That is for applications and
> systems to determine. Its just the same thing with IP 
> addresses. RFC 791
> specifies no methods for saying that two different IP addresses denote
> the same host. That is something that's left for things like 
> /etc/hosts
> files and DNS. RFC 2396 is the same way, it makes no statements about
> co-denotation since that is something that is left to things like RDF
> and OWL. 

Whether asserted or not asserted in 2369, what I was taking issue
with was what you nevertheless asserted -- that lexically distinct
URIs denote different resources.

If we presume (and I think we can and must) that the different layers
of the architectural cake have a consistent definition of what a
resource is, then such an assertion is in direct conflict with the
higher layers.

So if you consider RFC 2369 as defining resources in a way that
is not equivalent to that presumed by the higher layers, then we
have a problem -- either 

(a) RFC 2369 is too vague and allows valid interpretations which
    directly conflict with that of RDF/OWL, or

(b) RFC 2369 is too constrained and precludes interpretations
    which are required by RDF/OWL.

In either case, the definition of 'resource' in 2369 needs revision,
or else those higher layers will need to use something other than

Given that the primary (almost exclusive) "customer" of 2369 are
Web and SW applications, it would seem rather strange for 2369
not to facilitate a harmonious abstraction of 'resource' by all
of the layers.

> URIs won't solve that problem for them, but it doesn't get in
> their way of solving it either.
> > > ...
> > > > But you seem to be asserting that co-denotation is prohibited,
> > > > not simply possible. There is nothing left for OWL or other
> > > > layers to contribute, without coming into direct conflict with
> > > > what you appear to be asserting, that lexically distinct URIs
> > > > always denote distinct resources.
> > > 
> > > Its prohibited _at that layer_ by the simple fact that you have no
> > > language in which to express any of those concepts. 
> > 
> > Sorry. No. Just because you don't have the language to express
> > something does not mean it is either impossible or prohibited.
> Its impossible at that layer. Do it at a higher one.

I meant, it's not impossible in reality.

It's one thing to say the resolution is too low to discern some
feature and therefore that feature must be addressed at a higher
resolution (higher layer). It's quite something else to say
that if a feature is not discernable at a particular resolution,
that it is an impossiblity at any resolution.

> > It simply means you can't tell.
> > 
> > It's about granularity. RFC 2396 does not provide sufficient
> > granularity of specification to clarify co-denotation of 
> > resources. But that doesn't mean it is prohibited.
> Correct.....

But your assertion was that, since RFC 2396 cannot equate
the denotation of two lexically distinct URIs, then those
URIs *have* distinct denotation (as opposed to might have,
or could have, or can be presumed at this level to have, etc.)

Again, perhaps you didn't meant to make so strong an assertion,
but you did.

> > That's like saying that atomic structures are prohibited because
> > at the "layer" of my eyes, I can't see them.
> I didn't get that last analogy but the statement its based on is
> correct. co-denotation is something that RDF does to 
> RDF-resources, not
> URI-resources.....

It seems we have incompatable metaphysics.

Resources are just things in the universe of discourse. The fact
that URIs alone cannot provide information about their denotation
does not constrain their denotation, nor does it demand two
entirely different classes of resource.

RFC 2396 tells you about URIs. Not about resources. It mentions
resources because the whole point of URIs are to serve as a
means to denote resources. But all RFC 2396 can say is that
a given URI denotes some particular resource (thing) and 
whether two URIs are lexically equivalent or not.

But the fact that 2396 says nothing about denotational equivalence
does not mean the resources denoted by the URIs it defines are 
different from the resources that REST or RDF or OWL talk about.

Received on Wednesday, 16 July 2003 05:15:16 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:25:06 UTC