RE: ungetable http URIs

-------- Original Message --------
> From: Butler, Mark <>
> Date: 19 November 2003 15:48
> 
> > 
> > 3/ (Not quite sure what exactly you mean by "metadata
> > instance"
> 
> something like this
> 
> <http://web.mit.edu/simile/metadata/artstor/id#UCSD_41822000037356>
>       a       vra:Image ;
>       art:geographic
>       <http://web.mit.edu/simile/metadata/artstor/geographic#Spain> ;
        ((some RDF that got broken in email))

That is some RDF statements about
<http://web.mit.edu/....id#UCSD_41822000037356>.
There is no grouping in that RDF - I wondered it the term "metadata
instance" was more about the statings of these statements

> 
> > but ...)
> >    like 2/ a GET returns a representation.  If there is a
> > document/representation
> 
> All this information is in the RDFied Artstor corpus, so it sounds like
> the solution is to offer a HTTP interface to the Arstor corpus using
> the URLs used in the corpus. So my next questions are:  
> 
> 1. To do this I think we need to produce appropriate subgraphs, perhaps
> using a technique similiar to Alex Barnell's work on RDF objects,
> right?  
> 
> 2. Does Joseki support this - is it just a matter of publishing the
> data in Joseki? Or is it something for future work? 

Yes.  There is a "fetch" operation (its another query langauge)

	GET
http://simile.dspace.org/someCombineCorpus?lang=fetch&r=http://.....7356

	ie. KB.fetch(URI)

means to return all the relevant information about the resource identified
by the URI.  The determination of "relevant information" is a decision by
the server-side and different KBs can have different handlers - the client
then looks in the RDF to see what they got.  c.f. URIQA and TA.

It is like Alex's RDF object but is more server side.  A common handler is
the bNode closure from the identified resource as subject.

> 
> 3. If (2) is yes, does this have any implications for the URLs - do we
> need to change them so we can use Joseki? 

No - Joseki separates the KB URL from the resource in the KB.  This allows
any KB to have statements about any resource.  3rd part metadata.

I try to get round to replying to Kevin's Joseki email and give an overview.

> 4. We've agreed to keep the Artstor data private, so we can't expose
> artstor via HTTP. So does this mean we just use URNs? Making modelling
> decisions due to non-technical issues feels wrong?  

That is not a technical problem - security is orthogonal.  There are many
URLs I can't get at - oether people can.

> 
> Dr Mark H. Butler
> Research Scientist                HP Labs Bristol
> mark-h_butler@hp.com
> Internet: http://www-uk.hpl.hp.com/people/marbut/