- From: Graham Klyne <GK@ninebynine.org>
- Date: Mon, 28 Apr 2003 13:03:50 +0100
- To: Michael Mealling <michael@neonym.net>
- Cc: uri@w3c.org
At 12:01 27/04/2003 -0400, Michael Mealling wrote: >On Sat, 2003-04-26 at 19:25, pat hayes wrote: > > >First, > > >> note that the _semantics_ of a URI are not defined in > > >> this specification. Each URI scheme itself defines the > > >> relationship between URIs of that scheme and the resources > > >> they identify. In all known cases, that definition isn't > > >> enough to allow the URIs of that scheme to be used, by > > >> themselves, as unambiguous identifiers when trying to > > > > make logical assertions. > > > > Again, the problem is not that the logic requires unambiguity; quite > > the reverse, in fact: it is that imposing unambiguity as a defining > > characteristic of URIs is logically incoherent. > > > > Suggested modification, without the logic-bashing: > > > > > What is a Resource? Can a URI be used to identify a Concept? > > > > > This specification does not define the word 'resource' carefully, > > > nor does it define how a URI can be used to 'Identify' a > > > 'Resource'. The _semantics_ of a URI are not defined in > > > this specification. Each URI scheme itself defines the > > > relationship between URIs of that scheme and the resources > > > they identify or refer to. > > > > The reason for the last three words is that it doesn't make sense to > > say that a URI identifies a *single* thing, in almost any logical > > language. (In fact, it doesn't make much sense in any language, IMO, > > but leave that aside for now.) > >Pat, > I think that's one of the places that we're falling down on our >explanation. You assert that it doesn't make sense "in almost any >logical language". And that's the point, URIs aren't _in_ a logical >language, they exist outside all of them. URIs may be used _by_ a >language but at that point it is incumbent on the language/system that >uses them to define what it means concerning the things that URIs are >allowed to identify. > > As Graham has pointed out before, the problem here is that RDF, SW, >OWL, etc have erroneously attempted to inherit RFC 2396's concept of >Resources unmodified, and that was a severe mistake. IMHO, it is >probably worth inserting language into the new document suggesting that >it is extremely dangerous to simply inherit this definition of a >Resource without some sort of system specific profile of it. Systems can >do it but they have to be very clear that they are inherit it, not the >other way around... > > To simply inherit this documents concept of a Resource would be the >network equivalent of simply specifying an application layer (8) >protocol directly on top of IP, ignoring its admonitions to use things >like TCP/UDP/ICMP, etc that constrain IP to something that is usable by >higher layers. I don't recall saying that which you attribute to me. But it seems to me you are pushing much the same thing as Pat, in a different way. That the URI spec shouldn't be committing layered logical languages to specific interpretations of URIs. Where I (maybe) differ slightly from your line is that I think it's OK for the layered specifications to inherit the RFC2396 concept of a resource, but that RFC2396 should not say anything about a resource that it really does not need to say, so that there's less to cause problems for the layered specification. Running with your example of IP, it's important that IP *does* specify the syntax and interpretation of the IP header, but that it says as little as possible about the IP payload. One could imagine a (poor) specification that says the payload should start with something like a TCP/UDP header, providing some groundwork for most of the traffic that is actually conveyed by IP. But that would be entirely unhelpful for, say, ICMP. I think an important point that the URI specification probably does need to convey is that a resource is not the same thing as a representation of same. And enough background information that protocol developers use URIs in sensible ways. But beyond that, the nature of a resource should be constrained as little as possible by the URI spec. #g ------------------- Graham Klyne <GK@NineByNine.org> PGP: 0FAA 69FF C083 000B A2E9 A131 01B9 1C7A DBCA CB5E
Received on Monday, 28 April 2003 10:06:12 UTC