Re: The "info" URI Scheme

On 2003-10-02 16:05, "ext Karl Dubost" <karl@w3.org> wrote:

> 
> 
> Le mercredi, 1 oct 2003, à 12:26 America/Montreal, Hammond, Tony
> (ELSLON) a écrit :
>> The same thing with URI dereferenceability: URIs do not have to be
>> dereferenceable: Get over it!
> 
> Quite true... But if there's a possibility of deferenceability of URLs
> in URI. Some people will do. As you I don't get the necessity of info:
> scheme, specifically if you compare it to urn:
> 
> My problem with deferenceable URIs in the semantic Web is that it could
> cause harm to the whole semantic Web in some circumstances. I would
> rather prefer they are not deferenceable alltogether and at all.
> 
> <URI> is a label, reference, etc... (choose a name) which is generic
> for the use and the context of your work or community. The problem is
> that you have some URIs like http URLs which are deferenceable, and
> gives the possibility to have something at the end of it.

So what? The denotation of a URI has nothing to do with what
representations might be available via HTTP or any other protocol.

Whether there is or is not anything "at the end of it" has no
affect on reasoners using that URI.

> http://[domain name]/whatever/path/
> 
> the domain name can be bought by another party, the domain name is
> owned by someone, etc.
> 
> Why it's a problem?
> 
> Practices: Some people might find useful to put something at the end,
> for example we can imagine that the Web community decides to put
> documents (in RDF) that will make sense at the end of this URL. If the
> practice spread, it can be misussed and/or abused. Someone using an
> ontology will have suddenly a software that will not work because the
> expected thing at the end has changed, even if the practice is not in
> the spec.

That's normal life on the SW. Agent's beware.

> (I remind that tables in HTML were not created for layout,
> but it has been the case). So even in the spec, if you say that URIs
> MUST not be deferenceable, some people will do.
> 
> I would rather prefer a system, where URIs for semantic Web are not
> tied to a deferenceable object.

Then it's not the Semantic Web! It's just some closed expert system
operating in a bubble.

> A kind of example is for "words" in natural language (Semantic Web
> URI), they are not owned by people, so they are not dependent in their
> use from someone (except when using TM brands as common names) BUT you
> can find definition of these words in multiple dictionaries
> (Ontologies) private, commercial or public.

Err... and so different sources will assert different things about
particular resources, and one can choose which sources they trust
and which they don't. This is precisely how the SW works.

One can refer to Merriam Webster/"dog" or Collins/"dog" etc.
and those are distinct identifiers, specific to a given dictionary
(or even edition of a dictionary). Perhaps they denote the same thing,
but they also may have descriptions that differ or even conflict.

The SW cannot and will not be a closed, controlled space. And
because it will grow and change rapidly and in unexpected ways,
the SW architecture must be as flexible and scalable as the SW
will be dynamic -- just as the Web architecture is, which is why
the Web is such a success.

Patrick

Received on Thursday, 2 October 2003 09:54:07 UTC