W3C home > Mailing lists > Public > public-sw-meaning@w3.org > December 2003

RE: URI-meaning through collaboration

From: Bernard Vatant <bernard.vatant@mondeca.com>
Date: Tue, 23 Dec 2003 15:42:39 +0100
To: <algermissen@acm.org>
Cc: <public-sw-meaning@w3.org>
Message-ID: <GOEIKOOAMJONEFCANOKCMEMBCOAA.bernard.vatant@mondeca.com>

> > > I think I finally understood the "meaning of URI" issue. I
> > > am very curious what people think of the following:
> > >
> > > "The meaning[1] of a URI is the sum of the semantics of all
> > >  uses of that URI".

This goes somehow along the lines of a paper I presented two years ago

Although my thoughts on this issue have somehow evolved since, I still
agree with most of it. The difference with what you suggest is that IMO the
meaning of some subject (identified by its URI or otherwise) should be the
result of some actual *conversation* about it. A conversation means a
community of speakers eager to set a consensus. If anyone is asserting
anything about the subject without caring about consensus, I agree with
Peter that could lead to "denial of service" attacks on the semantics of
the subject. I could spider the Web, and for a:anyResource I'm not happy
with, assert:

a:anyResource	owl:sameAs		owl:Nothing

This is playing the bad guy indeed, but quite easy to do. In an OWL
environment, it could trigger large-scale inconsistency. In a Topic Map
environment, it could trigger "semantic collapse" of huge topic maps to a
single black-hole topic.

In fact, one could define a community of use(rs) by a shared meaning,
represented in a set of ontologies making non-contradictory assertions
about some URI. Which means - I'm again making that point, sorry about
that - there is and can't be *absolute* and *universal* meaning of an URI,
more than of any other kind of name or identifier. But one can define a
scope of shared meaning, based on consensus and best practice. Now how to
detect the bad guy inthere is a pending issue ...

> > I don't think that this is particularly useful.  What happens
> > if a URI is used in contradictory fashion?  Does this make *the*
> > of the URI be a contradiction?

If we ask for *the* unique, global, absolute, meaning, this will always
lead to contradiction, not to say religious wars. The only way to avoid
that is a process enabling a community to share some meaning, and let
outsiders share another one. The way it goes with every language in the
world, AFAIK. How technically one can do that except behind the firewalls
of one's intranet, I really don't know.

> Certainly not.
> If you don't know the meaning of a word....do you ask an
> authority or do you look it up in a dictionary?

First I try to figure out what whoever is using the word actually means. Of
course if whoever is referring to some authority, I'm better off. Assuming
(s)he is actually consistent with what the authority means ...

> [Since natural language dictionaries reflect the current common use of
> the words of a language, wouldn't it be nice to have URI-dictionary
> services on the Web?]

Hmmm ... is not that called Published Subject Indicators?

> > > The main idea here is collaboration. Each use of a URI contributes
> > > to it's meaning and the (current) meaning is the sum of all
> > > such contributions (known to date).

All the contributors in a community of consensus, yes. *All* the
contributors, certainly not ...

> > Well, this would certainly lead to a nice denial of service
> > attack on the Semantic Web.  Just use lots of URIs in unusual ways,
> > polluting *the* meaning of these URIs.

Agreed. See above

> Either way you gotta be careful which (RDF-)documents you trust
> and which not. Or do you suggest to only use those issued and
> controlled by the authority? What if an under-paid employer that
> happens to be the Web master goes nuts and publishes millions of
> of contradictory RDF statements?
> So while this is definitely a semantic Web issue, it is not
> limited to my suggestion.

Agreed also. It is a generic issue in all semantic applications. I would
say it's a generic issue with language. You need semantic regulation in
every conversation, and shared semantics are always the result of some
negociation. Remember Sarah and Jim discussing what a planet can be. [1]

> > > This creates a picture of the meaning of a URI being in constant
> > > flow, but gaining stability through increased (similar) usage. If
> > > a URI does not reach a critical point of stability...well, then it
> > > propably wasn't good enough in the first place.

> > Hmm.  Perhaps a different metric would be useful.  If a URI
> > does not reach contradictory status then it probably wasn't good enough
> > the first place.

Sure. For Jan who seems to miss the point, I guess Peter means that if you
don't meet someone who does not agree with what you say, it's because
either you did not say much, or you did say something fuzzy enough to be
agreed upon by anyone at no real cost (which amounts to the same,
basically). IOW, a good semantic definition has some arbitrary content that
one can clearly disagree with.

> > > This puts the naming authority in a position of responsibility
> > > to care for a young and fragile URI, slowly raising it to be
> > > strong (semantically stable) as opposed to 'dictating it's
> > > semantics up front).

I like this in principle, as said above, also I don't figure technically
how it could be achieved.

> > One nice aspect of this theory of meaning is that the naming
> > authority has no special powers.

Yes. "Big Elm Cross" remains "Big Elm Cross" centuries after the last elm
is dead, and even remembrance of what an elm could look like has faded
away, and all you have there now is a suburban highway interchange ...

> > > After all, who can 'define' the meaning of a name if not the
> > > community that uses the name?

Making this happen in the Semantic Web would mean reconciliating formal
semantics and real-world conversation. Seems like mixing oil and water,



[1] Jack Park's "XML Topic Maps" ISBN 0-201-74960-2
see Chapter 5 : "Conversation, Names and Subject Indicators"
Received on Tuesday, 23 December 2003 09:42:57 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:42:15 GMT