RE: DESCRIBE ENDPOINT



> -----Original Message-----
> From: public-rdf-dawg-request@w3.org [mailto:public-rdf-dawg-request@w3.org]
> On Behalf Of Lee Feigenbaum
> Sent: 01 July 2009 07:50
> To: Simon Schenk
> Cc: Steve Harris; Axel Polleres; 'RDF Data Access Working Group'
> Subject: Re: DESCRIBE ENDPOINT
> 
> Simon Schenk wrote:
> > Am Dienstag, den 30.06.2009, 21:58 +0100 schrieb Steve Harris:
> >> <> Might be more conventional. That depends on what BASE defaults to
> >> though. It will still map to some graph inside the store, which
> >> restricts the things you can import into the store without confusing
> >> the discovery.
> >
> > I don't understand. If <> is mapped to some graph, whose name no one
> > except for the endpoint itself needs to know, the name does not need a
> > globally valid name. Then, there should be no problem?
> 
> I believe the problem is that <> has a well-defined name - it's whatever
> base URI is in effect when executing the query.
> 
> Don't ask me to repeat/recite the rules for determining the base URI
> though, because I always have to go look them up.
> 
> Or maybe I'm misunderstanding :)
> 
> Lee


Exactly right - it's a relative URI in the syntax and on parsing the query will resolve the URI against the base.  C.f. Use of <> in Turtle.

There is big difference between 

DESCRIBE SELF
And
DESCRIBE <uriThing>

Because <uriThing> is (supposed to be) globally unique after resolution (except if localhost is involved.

DESCRIBE SELF is relative to the receiving endpoint.

Any system that uses well known URIs to denote things or special graphs should really cope with the fact that the well known URI is web-wide and should not name one thing on one system and another on another system.

 Andy

Received on Friday, 3 July 2009 19:18:41 UTC