W3C home > Mailing lists > Public > public-xg-geo@w3.org > August 2006

Re: INSEE releases OWL ontology and RDF data for geographical entities

From: Eric van der Vlist <vdv@dyomedea.com>
Date: Sun, 06 Aug 2006 17:16:06 +0200
To: Joshua Tauberer <jt@occams.info>
Cc: Hans Teijgeler <hans.teijgeler@quicknet.nl>, 'Xiaoshu Wang' <wangxiao@musc.edu>, 'Dan Connolly' <connolly@w3.org>, semantic-web@w3.org, public-xg-geo@w3.org
Message-Id: <1154877367.21743.15.camel@localhost>
Le dimanche 06 août 2006 à 10:59 -0400, Joshua Tauberer a écrit :
> Eric van der Vlist wrote:
> > Le dimanche 06 août 2006 à 15:45 +0200, Hans Teijgeler a écrit : 
> >> Please elaborate on what you exactly mean with that roundtrip.
> > 
> > If I understand the spec correctly, if you load a XML/RDF document with
> > weird namespaces names such as http://foo.com/b or http://foo.com#b in a
> > RDF parser and serialise the model back into XML/RDF, in addition to the
> > common "roundtrip" issues, the result could use different namespace and
> > local names than the original document!
> 
> What difference would that make (besides it possibly not looking as
> pretty)?  The input and output documents will still be entirely equivalent.

Please read my mail in the context of the question to which it is
answering!

In that specific case (see my answer to Hans), I don't think we can say
that <ar xmlns="http://foo.com/b"/> and <bar
xmlns="http://foo.com/"/> are really equivalent even if they generate
the same triples.

> Just to throw in my two cents into this thread, while I'm emailing
> (although I don't expect to say anything no one has said before) --
> 
> It seems like issues about # versus /, choosing good namespace URIs, and
> dereferencing are all very fragile issues.  As we are right now, we've
> been told (by the specs, implicitly if not explicitly) that there's
> nothing in a name.  They're supposed to be opaque.  So if I start
> coining millions of http:-schemed names, I have an expectation that
> tools *are not* going to bombard my server with millions of HTTP GETs
> trying to find things because you're not *supposed* to look inside a URI.
> 
> If that's not what we want, that is, if there is any (formal) level of
> expectation that http-URIs are intended both 1) to denote an entity and
> 2) to be potentially dereferencable to something, then the W3C better
> say so very soon so I know that if I don't want millions of HTTP GETs, I
> better not use http:.
> 
> In the meanwhile, until some standard says that http: URIs are expected
> to be even potentially dereferencable, my feeling is that we should just
> drop the idea that we can dereference anything unless some triple
> explicitly says so.

Here, we're back to the starting point of our debate!

That's what we thought too and that's the principle we've followed with
the current version of the INSEE OWL ontology.

Having been told that this was no longer a subject of discussion, that
this was not how the Semantic Web worked and that we were violating the
rules of the Web Architecture, I am now trying to find a solution where
we do not rigidly tight the identifiers of abstract resources to the
addresses of web resources and can still let people dereference our URIs
if they really want to :) ...

Eric 

> (OTOH, I have been using tag: anyway because I find http: unnecessarily
> confusing.)
> 
-- 
GPG-PGP: 2A528005
Freelance consulting and training.
                                            http://dyomedea.com/english/
------------------------------------------------------------------------
Eric van der Vlist       http://xmlfr.org            http://dyomedea.com
(ISO) RELAX NG   ISBN:0-596-00421-4 http://oreilly.com/catalog/relax
(W3C) XML Schema ISBN:0-596-00252-1 http://oreilly.com/catalog/xmlschema
------------------------------------------------------------------------

Received on Sunday, 6 August 2006 15:16:25 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:21:18 GMT