RE: Comment on "Meaning and the Semantic Web"

> From: Bijan Parsia
> Sent: Tuesday, June 01, 2004 10:02 PM
> To: John Black
> 
> On Jun 1, 2004, at 9:12 PM, John Black wrote:
> 
> >> -----Original Message-----
> >> From: Dan Brickley [mailto:danbri@w3.org]
> >> Sent: Wednesday, May 05, 2004 4:10 AM
> >> To: John Black
> >> Cc: public-sw-meaning@w3.org
> >> Subject: Re: Comment on "Meaning and the Semantic Web"
> [snip]
> >> For example, I can, it is commonly supposed, change the 
> OWL/RDFS/HTML
> >> at http://xmlns.com/foaf/0.1/ and alter the (ahem) meaning of
> >> thousands (millions, actually) of RDF/XML documents. By 
> editing a few
> >> files and typing 'make site', I can make docs that weren't 
> currently
> >> true of the world, true of the world; and vice-versa.
> >
> > But this is true if you owl:import http://xmlns.com/foaf/0.1/ also,
> > isn't it?
> 
> So, using owl:imports introduces a dependancy that just using 
> the terms 
> doesn't.
> 
> You could also cut and paste the FOAF ontology into your document by 
> hand. That would, on our view, preserve you from not-in-your-control 
> meaning shifts.

Agreed, this would do it.  But it seems to violate something that has 
been essential to the success of the web architecture.

> You could also import http://foo.com/myfoafmirror/ (pace copyright 
> issues) (which actually might come up with owl:imports). The imported 
> document could use the foaf prefix and provide more stabilty.
 
You mean make a mirror of a snapshot of the FOAF site? a cached version?
and import that? I don't know what you mean.

> >  So the imports-only proposal offers no defense against this.
> 
> Sure it does. See above. It may
> 
> > The meaning of documents under this proposal is just as 
> dependent on 
> > the
> > stability of the ontological context it imports as other 
> documents are
> > dependent on the stability of the schemas whose namespaces they 
> > include.
> 
> Except that imports and URI use are separated. So you have more 
> flexibility. We didn't claim it was a *grand* solution, only 
> one with a 
> bit of critical flexibility.
> 
> >> Since I'm a reasonably responsible person, this is 
> probably not a big
> >> deal. I promise to be good. But try explaining that to someone
> >> building an ecommerce system that depends upon  
> externally-managed RDF
> >> vocabularies.
> >
> > Right.  And now that you have explained it, I will refrain 
> from using 
> > FOAF
> > for anything serious until I have a way to make sure my 
> documents are
> > interpreted using the exact version of the FOAF spec that 
> existed when
> > I wrote them using that namespace or when I imported it.
> 
> In my view, you have it now. Cut and paste, or mirror and import. Are 
> this horky, sure. But at least they, in fact, work. This is 
> all we were 
> advocating.
>
> >  Personally,
> > I predict that you *WILL* change it at some point in a way 
> that changes
> > the meaning of millions of statements. And this is the 
> point Tim was 
> > trying
> > to make, I believe, when he said "that use of a URI in RDF implies a
> > commitment to its ontology."
> 
> What? Your prediction is Tim's point? Sorry, lost me.

Ok. Poorly stated.  New paragraph. It reminds me of Tim's point...

> 
> > http://lists.w3.org/Archives/Public/www-tag/2003Jul/0022.html For, 
> > just as
> > changing the FOAF spec. would alter the sense of millions 
> of documents,
> > failure to use the correct FOAF spec., when interpreting some RDF 
> > document
> > dependent upon it, would alter the sense of the document.
> 
> Still lost, sorry.
> 
> Hmm. I *think* you mean that, sans import, if I fail to (in 
> effect) add 
> the imports in, then I've changed the meaning of the 
> document? Piffle. 
> Not *my* documents. Refraining from an imports seems to me to be a 
> powerful way to signal what I mean by foaf:knows (*not* in 
> the Biblical 
> sense).
> 
> This is a difficult point for a number of reasons. For 
> example, part of 
> the point of moving to a machine processible formalism is so that the 
> *formalism* can encode the "specification". But our specs aren't 
> separate from the things they "specify" -- that is, it's very much 
> unlike the relationship between a grammar and the strings recognized.
> 
> Oh wait, I see. Sure, when interpreting *some RDF document dependent 
> upon the correct FOAF spec* (for its meaning). But, sans imports, 
> there's no way in rdf to indicate that dependancy. And that's 
> fine. RDF 
> meaning is perfectly reasonable. If I want to conform to some other, 
> tighter spec, there are any number of ways to indicated that.
> 
> >   If it is "good"
> > of you to refrain from changing the spec. in this way, it seems it 
> > would
> > be similarly good of you to use the spec., that is commit 
> to it, when
> > interpreting someone's URIs.
> 
> Hardly. In one case *you* affect the meaning of *my* 
> documents. In the 
> other, my different use doesn't affect the meaning of *your* uses. If 
> the difference, thus put, isn't brutally obviously, then I'm at a 
> complete loss.
> 
> >   The scale is different but the effect is
> > the same.
> 
> Clearly not.
> 
> >   By refusing to respect the ontological context of a document,
> [snip]
> 
> You are imposing this very specific notion of ontological context. It 
> happens not to be in any spec, nor is it, contra some view, at all 
> imposed by web architecture. You can propose it, but I, for one, will 
> vigorously resist.
> 
> But I feel like there must be a thinko here. I'm finding this 
> symmetry 
> argument impossible to take seriously.

Let's use a little stage setting here.  Suppose I have a search engine 
agent that populates my web site with results of a semantic web enabled 
search.  I am in particular looking for FOAF documents with a particular 
content.  Now everything is working fine until one day, documents that 
used to be retrieved correctly stop showing up and others that I've never 
seen before - and don't want - do start showing up. Now I have to 
debug this. Two equally plausible hypotheses occur to me. Either Dan 
has gone bad, and decided to mess with the FOAF schema, or my search 
agent has failed and its mistakes are equivalent to what Dan might 
have done.  I can't tell which.  My documents, using the output 
of the search engine, are defective in the same way in either case.
This is the symmetry argument.  In a society of communicators, 
correct interpretation is just as important as correct publishing.


> 
> Cheers,
> Bijan Parsia.
> 
>

Received on Tuesday, 1 June 2004 23:14:31 UTC