- From: Leigh Dodds <ldodds@ingenta.com>
- Date: Tue, 16 May 2000 19:27:42 +0100
- To: xml-uri@w3.org
> Are there any use cases for namespaces where relative URIs are actually > necessary? The only case where I can see relative URIs might be useful is when you: 1. Want the Namespace identifier to point to a resource (possibly a schema) 2. Want to distribute that resource with the document. As you know relative URIs are normally used to facilitate the movement of webpages - as the absolute URLs aren't 'hardcoded' into the HTML. It strikes me that this what relative namespace URIs are about. Its not about tying up with semantics, its about resource resolution/management. Thinking along these lines (which may be incorrect, granted, but run with me a ways :) led me through the following chain of thought: 1. We need to name things. 2. A name is a useful handle to hang semantics off. (I know these first two points cover a lot of philosophical debate). i.e. you have to identify what you're describing. 3. My understanding of TimBLs vision (the Semantic Web) is that the namespace identifiers will ultimately point to 'something' (schema/resource/RDf desc.). And hence to: A. An absolute URI makes it easy to resolve the identifer to 'something'. B. A relative URI makes it easy to manage your 'something's, as you can move them around (and just redefine the base uri). It also allows you to use a dereference locally rather than across the 'net. I'm not talking about absolutizing, but rather dereferencing the identifier to that 'something' we want to fetch. C. The base URI should not be defined by *location* it should be defined by something else (XML Base?, or something else, its not clear if this is well-specified as yet) - this solves the issue of a document move causing a semantic change, but doesn't address the interoperability problem (which you raised earlier in the discussion). D. The need for relative URIs may go away if there was another solution for this entity/resource management problem. Do we need a catalog type mechanism to allow the flexibility of relative URIs? Hopefully that makes some kind of sense. I guess like Simon I'm trying to get to the heart of why someone wants to use relative URIs. The root cause of the problem to me seems to be whether you want the NS identifiers to point to a real resource. If you do then a relative URI gives you some benefits. I'm looking forward to having my chain of thought derailed1 :p L.
Received on Tuesday, 16 May 2000 14:37:51 UTC