W3C home > Mailing lists > Public > www-rdf-interest@w3.org > October 2002

Re: Meaning of URIRefs (No Ancient Philosophical Disputes!)

From: Sandro Hawke <sandro@w3.org>
Date: Fri, 25 Oct 2002 18:25:40 -0400
Message-Id: <200210252225.g9PMPeE26447@wadimousa.hawke.org>
To: Dan Brickley <danbri@w3.org>
cc: www-rdf-interest@w3.org


> > My definition of "definition" does *not* require settling any
> > philosophical disputes.  I'm just saying that when someone makes up a
> > name for something, they get the chance to make certain assertions
> > about that thing.  The choice to use a name involves accepting the
> > truth of those assertions.
> 
> That makes disagreement difficult. If I can't use your terms without the
> baggage that comes with them, I'll find it hard to use RDF to express the
> detail of certain kinds of disagreement.

As I've said in this thread several times before today :-) if you want
to do this kind of thing you need to either (1) quote the URI, (2)
make up your own URI or bNodes for the same object.    It's like when
I want to talk about PFPS's term "George-Bush-the-lesser" without
making any aspersions on the President.

> A minimal consequence of this is that it suggests RDF reification is
> broken, since I can't even "scare quote" something you say without using
> your term and (on your reading) endorsing the package of assertions it was
> introduced to the Web with.

RDF Reification is broken, aside from all this.  The URIRefs need to
go inside quotes (strings), because otherwise reasoners are likely to
do evil (er, unsound) things with them. 

> > This is in spirit of what the editor's draft of the Concepts document
> > already says -- I'm just trying to make sure people understand the
> > complete ramificiations of this view and trying to find engineering
> > ways of making those ramifications acceptable.  One of the most
> > unfortunate ramifications is that term-reusers have to accept all the
> > assertions in the "definition" whether they are "definitional" or not,
> > because we can't really distinguish between the two.  This recent
> > thread is about whether that consequence is fatal or not.  I'm arguing
> > it's not all that bad.
> 
> If the RDF Concepts doc really says this, it's broken and needs fixing.
> URIs shouldn't carry secret baggage. Consider a bookmark system that
> mentions lots of URIs, or something like DMoz / Open Directory. If I
> wanted to protect my URIs from being 'talked about' by others, I could
> just define them with lots of libellous (or racist, sexist or otherwise
> offensive) claims such that nobody else would ever dare to mention them.
> That doesn't seem optimal...

In some views it's only when used as predicates that URIRefs carry
baggage.   Maybe that's a middle ground here.

In fact, a bookmark system probably SHOULD quote URIs, though, since
it doesn't want them to be reasoned about as symbols, and does want
their text to be handy to feed to users.   Also, such URIs might
syntactically be URIRefs (if they are bookmarks to HTML IDs), which
could cause all sorts of logical complications.  Ugh.

     -- sandro
Received on Friday, 25 October 2002 18:26:22 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:51:56 GMT