Re: MISC: Internet Media Type registration: proposed TAG finding

>On Fri, 2002-05-24 at 23:05, Peter F. Patel-Schneider wrote:
>>  From: Dan Connolly <connolly@w3.org>
>>  Subject: Re: MISC: Internet Media Type registration: proposed TAG finding
>>  Date: 24 May 2002 21:34:31 -0500
>>
>>  > On Fri, 2002-05-24 at 19:09, Peter F. Patel-Schneider wrote:
>>  > > From: Dan Connolly <connolly@w3.org>
>>  > > Subject: Re: MISC: Internet Media Type registration: proposed 
>>TAG finding
>>  > > Date: 24 May 2002 16:13:32 -0500
>>  > [...]
>  > > > > The author of the best-friend document, by choosing to use
>>  > > > ont:UniqueProperty class, licensed inferences
>>  > > > based on the specification of that class. The conclusion
>>  > > > that "35" is an :age of :margaret is supported
>>  > > > by the DAML+OIL spec.
>>  > >
>>  > > Yes, but not by the RDF spec, and any agent has no business labelling
>>  > > anything as RDF inferences that are not sanctioned by the RDF 
>>model theory.
>>  >
>>  > Hmm... that's one way to think of 'RDF inferences'.
>>  > It's not one that appeals to me.
>  > >
>>  > > > I can think of two agents (cwm and Euler) that
>>  > > > do a lot more than simple entailment, when
>>  > > > asked to. I think of them as RDF agents.
>>  > >
>>  > > They are not.
>>  >
>>  > Er... I accept that as your opinion.
>>  > I disagree.
>>
>>  Then it is up to you to show how the RDF specification admits cwm and Euler
>>  as RDF agents.
>
>You seem to think it's reasonable for you to introduce a
>term (RDF agent) into the conversation, without defining it,
>then define it later, and then hold me to your definition
>as if I had accepted it.
>
>I repeat: I don't agree to your definition of 'RDF agent',
>i.e. something that computes inferences that are
>valid per the RDF model theory, and no others.

Evidently we have a serious breakdown in communication here. I have 
always understood terminology like 'RDF agent' or 'RDF engine' in 
more or less Peter's sense. I wish I knew what your sense of the term 
means.

>  >  I claim that they are not, and back up my claim by your
>>  statement that cwm and Euler do more than is santioned by the RDF model
>>  theory.
>
>When I first saw the term 'RDF agent' in this thread, I
>inferred that it meant something that would draw all
>conclusions that are sanctioned by the RDF model
>theory, plus any others it was instructed to make.
>By that definition, cwm and Euler are
>RDF agents.

Unfortunately, so is almost any reasoner on the planet, since RDF is 
such a tiny subset of first-order logic. So the concept of 'RDF 
agent' that you are using seems almost vacuous. Also, it seems to 
have nothing particularly to do with RDF, so I am at a loss to 
understand why you find it a useful notion. On this meaning, for 
example, something that drew conclusions that were actually denied by 
the RDF model theory could count as an RDF agent.

>....
>>  and
>>  if they know that they should know, is being deceitful.  Anyone who calls
>>  the results RDF is verging on deceit.
>
>Hmm... cwm writes out its results in a format that conforms
>to the RDF syntax spec. That's not RDF?

It is encoded in RDF syntax, but that is trivial. Does its meaning 
conform to the RDF specs? No, for cwm (it uses arc labels to denote 
things that are not properties, like logical implication.) So I would 
say that it is not RDF, indeed. N3 is not RDF.

>I'm pretty sure
>I don't understand what you mean by 'calling the results RDF'.

It is RDF if it conforms to the RDF specs, which includes the model theory.

>.....
>My position on media types for the Resource Description Framework
>is that the RDF model theory isn't the end of the story for the
>whole framework; folks that send documents labelled
>with application/rdf+xml (or whatever) definitely license
>all simple entailments. But they also license other inferences,
>based on the vocabulary they choose to use in their document.

HOW do they licence those other inferences?? From where comes such a 
licence? The choice of vocabulary does not extend the class of valid 
RDF inferences. What does it mean to say that something is RDF, but 
one is expected to draw non-RDF-valid inferences from it? This is 
meaningless.

BTW, I wish that you (or someone) would articulate this 'whole 
framework' vision that you seem to have. Right now it seems to be 
part of the W3C secret lore, but isn't being vouchsafed to the rest 
of us. I know y'all are very busy, but peer review does have some 
advantages.

>So just because an agent's inferences aren't simple entailments
>doesn't mean they're false, wrong, deceitful, or unjustifyable.

Of course. But it does mean they are not sanctioned by the RDF specs, 
so an agent that does draw them is NOT conforming to the RDF specs, 
so is not an RDF agent (in any useful sense that I can determine.). 
If we adopt your convention for 'RDF agent', what is the utility of 
even bothering to publish a semantics for RDF? An RDF agent in your 
sense can apparently draw whatever conclusions it likes, so we might 
as well have saved ourselves a lot of work and just published an 
RDF/XML syntax, and let the world do whatever it likes with it. In 
fact, we could have saved even more work and just told everyone to 
use XML in whatever way they like.

>On the contrary: I suggest there's an emerging protocol for
>introducing new vocabularies into the framework, and that
>documents labelled with this media type are part of that protocol.

What emerging protocol?

>
>>  > > > > > > Sure, but the non-understood part may completely 
>>change the meaning of the
>>  > > > > > > other part,
>>  > > > > >
>>  > > > > > No, I don't think so...
>>  > > > >
>>  > > > > Why not?  What about
>>  > > > >
>>  > > > >	<fol:negation>
>>  > > > >	  <fol:statement>
>>  > > > >	    <rdf:Person rdf:about="#John">
>>  > > > >	      <loves rdf:resource="#Mary">
>>  > > > >	    </rdf:Person>
>>  > > > >	  </fol:statement>
>>  > > > >	</fol:negation>
>>  > > >
>>  > > > I'm not sure how to parse that. If I presume that fol:negation
>>  > > > is an RDF typednode, and convert to (something close to) KIF, I get:
>>  > > >
>>  > > >   (exists (?n)
>>  > > >     (rdf:type ?n fol:negation)
>>  > > >     (fol:statement ?n #John)
>>  > > >     (rdf:type #John rdf:Person)
>>  > > >     (loves #John #Mary)
>>  > > >   )
>>  > > >
>>  > > > so... ?n is a negation, and it's somehow related to John.
>>  > > > I think you meant to express something like
>>  > > >   (not (loves John Mary))
>>  > >
>>  > > That is indeed what I meant. I expressed it in RDF syntax, 
>>which according
>>  > > to you, should be perfectly fine.  I used my extension to RDF reasoning,
>>  > > which, according to you, should also be perfectly fine.
>  > >
>>  > I can't make sense of this extension. Maybe you could explain
>>  > how it works a little more generally than this one example.
>>  > I don't understand how the fol:negation class's specification
>>  > reaches out thru #John to the (loves #John #Mary) triple.
>>
>>  But according to your rules, any method of providing meaning for this
>>  extension is adequate.
>
>No, only those that are monotonic. i.e. when you specify an extension,
>you can only throw out models; you can't put any back in.

If we take that strictly, then 'layering' is impossible.

>  >  OK, the method that I will use is that it means
>>  what I mean it to mean. 
>
>So long as it's monotonic, then you haven't provided an example
>that ...
>
>>  > > > > > > [...] the non-understood part may completely change 
>>the meaning of the
>>  > > > > > > other part,
>
>
>>  This is no harder for an RDF-only agent to understand than the DMAL+OIL
>>  meaning of daml:UnambiguousProperty.
>>
>>  > In particular, I don't understand it well enough to
>>  > see if it's non-monotonic.
>  >
>>  Why should this be a problem?
>
>Because that's against 'my rules'; i.e. the rules I'm
>suggesting for this new media type.

1. What has the media type got to do with this discussion? (are we 
going to have different inferences when something is displayed on a 
TV screen from when it is printed on colored paper??)
2. If 'media type' just means 'according to one set of rules' , then 
Peter can define his own set of rules, which might allow nonmonotonic 
extensions.

>
>>  > > > but that's really awkward to do in RDF, but I suppose
>>  > > > it could look like:
>>  > > >
>>  > > >   <fol:Negation>
>>  > > >     <fol:predicateSymbol>http:...#loves</fol:predicateSymbol>
>  > > > >     <fol:subjectSymbol>http:...#John</fol:predicateSymbol>
>>  > > >     <fol:objectSymbol>http:...#loves</fol:objectSymbol>
>>  > > >   </fol:Negation>
>>  > > >
>>  > > > where fol:Negation is specified ala:
>>  > > >
>>  > > >   (forall (?n ?s ?p ?o)
>>  > > >     (<=>
>>  > > >       (and
>>  > > >         (predicateSymbol ?n ?p)
>>  > > >         (subjectSymbol ?n ?s)
>>  > > >         (objectSymbol ?n ?o)
>>  > > >         (wtr
>  > > > >            '(not (holds `(str->sym ?p) `(str->sym ?s) 
>`(str->sym ?o))) )
>>  > > >       )
>>  > > >       (rdf:type ?n fol:Negation)
>>  > > >   )
>>  > > >
>>  > > > But even in that case, I *only* eliminate models when I
>>  > > > learn the meaning of fol:Negation (i.e. when I add that axiom).
>>  > >
>>  > > Yes, and this is no longer RDF.
>>  >
>>  > I accept that as your opinion. I disagree.
>>  > Argument by assertion is really no fun.
>>
>>  Well, how could it be RDF?  It doesn't match any syntax for RDF.  QED
>
>
>Please allow me to clarify; I might publish, at the address
>of fol:Negation, something ala:
>
><rdfs:Property rdf:about="#specInKIF">
>   <rdfs:subPropertyOf rdf:resource="http:.../rdf-schema#comment">
></rdfs:Property>
>
><rdfs:Class rdf:about="#Negation">
>   <specInKIF>
>    (forall (?n ?s ?p ?o)
>      (<=>
>        (and
>         (predicateSymbol ?n ?p)
>          (subjectSymbol ?n ?s)
>          (objectSymbol ?n ?o)
>          (wtr
>             '(not (holds `(str->sym ?p) `(str->sym ?s) `(str->sym ?o))) )
>       )
>       (rdf:type ?n fol:Negation)
>   )
>   </specInKIF>
></rdfs:Class>

Sigh. Yes, you might. However, you have here written something in 
KIF, not RDF (and full KIF 3.0, I might add, including wtr and 
quasi-quotation, which goes way beyond simple FOL) . If you are 
willing to use KIF, why in God's name are we piddling about with 
RDF?? It would all be easier and simpler if it was written in KIF in 
the first place (in some XML-ish syntax, if you like) .

Pat
-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Thursday, 30 May 2002 14:12:59 UTC