Re: The 'resource' identified by a namespace name URI should be the namespace

On Mon, 5 Jun 2000, John Aldridge wrote:

> At 16:30 03/06/00 -0400, Dan Brickley wrote:
> >On Sat, 3 Jun 100, John Cowan wrote:
> > > Tim Berners-Lee scripsit:
> > >
> > > > It also prevents the namespace URI being usefully dereferenced,
> > > > as dereferencing it will get only the namespace name, and never any other
> > > > information about the namespace.
> > >
> > > Granted, but since there is no guarantee what you might get
> > > (an XML Schema? an RDF Schema? XLinks to various things? Human-readable
> > > HTML? Nothing at all?), this seems less than useful.  If you want
> > > to know things about a namespace, discover some RDF that tells
> > > you about it.
> >
> >I think 'less than useful' is a little unfair here, though RDF would be
> >one of my favourite things to find when dereferencing...
> 
>    :
> 
> >The thing that hangs together this (typically
> >weblike) heterogeneity is that whatever we find when we dereference
> >some namespace is authoritative. Whether I find HTML, MP3, a public key
> >(hmmm...) or XML/RDF when I dereference a namespace URI, I
> >should be able to treat that data as coming from the owner/manager of that
> >Web vocabulary, and therefore something that can tell me something of the
> >meaning of that vocabulary.
> 
> This works fine if...
> 
> (a) Only the creator of the namespace (i.e. the person with the ability to 
> place things at the namespace URI) should be able to make descriptive 
> statements about the namespace.
> 
> (b) A single formulation of that descriptive metadata is appropriate to 
> support all the various processing applications which the world will apply 
> to documents using names from the namespace.
> 
> I don't believe either of these.

Neither do I! Take the Dublin Core Metadata Initiative for example. You
might expect to find schemata using http://purl.org/dc/elements/1.1/
describing their vocabulary. Dozens of other people will want also to make
statements about those concepts (in Dublin Core land we have 26
non-english translations of the definitions into other languages). Lots of
other metadata systems will want to express mappings to, and rules about
the use of, the Dublin Core element set. 

I don't see a tension here. 

- the terms defined in a Web vocabulary need Web identifiers so others
  can make statements about them.

- we need to know which such statements are authoritative, ie. produced
  by the creator of that vocabulary, not gossip, mappings etc produced
  elsewhere.

URIs help with the former; connecting namespace URIs to notion of
authority helps with the latter.

In Dublin Core, the scenerio we seem to be moving towards is one in which
the DCMI 'usage committee' after-the-fact endorse the mappings and
annotations produced out in the community, eg. multi language annotations,
mappings into other schemas etc. For this we need Web identifiers for
everything, and some notion of who-says-what. Eg scenario: the namespace
URI might tell us a bunch of authoritative stuff about some vocab, including
digital signature related information that allows us to filter other
statements we later discover, such that we know which of those subsequent
statements were made by the namespace owner, and which were merely
gossip.

Dan

Received on Monday, 5 June 2000 12:57:48 UTC