W3C home > Mailing lists > Public > www-rdf-logic@w3.org > June 2001

Re: RDF Statements as floating Cons Cells

From: pat hayes <phayes@ai.uwf.edu>
Date: Fri, 8 Jun 2001 11:15:05 -0500
Message-Id: <v0421016cb746a3ae2fff@[205.160.76.219]>
To: Drew McDermott <drew.mcdermott@yale.edu>
Cc: www-rdf-logic@w3.org
>   [Pat Hayes]
>   Oh God, I am inclined to give up at this point. Why are we even
>   bothering to try to adapt this unbelievably broken system to make it
>   do something which it is  incapable of doing? I thought that I could
>   see a way to extend RDF to add more complex syntax to it, and now you
>   have convinced me that it can't be done.
>
>   ...
>   How does it stop someone adding
>
>   X3a X4b S
>   X4b nil T
>
>   and making the structure 'doubly defined' ? There is no way to do
>   that, as far as I can see. At this point I think I shall just go home
>   and go to bed.
>
>Well, I hope a good night's sleep has revived you.

With coffee, yes, thanks.

>I don't quite see why "double definition" is so scary here and not
>elsewhere.  We could have an axiom

Written in what notation?

>ruling it out, so that if the same
>list were defined twice we could conclude that the CAR in one
>definition must = the CAR in the other, and so forth.  We might get a
>contradiction out of it.  Or whatever.  But is it any different from
>father(fred, sally) and father(murderer(mary), sally)?  We don't
>conclude Sally has two fathers, but that fred = murderer(mary).

You are talking about reasoning in RDF (plus something that encodes 
axioms) *about* lists. I was thinking of this trick as a way to morph 
datastructures into RDF, not have RDF talk about datastructures. I 
was basically trying to trick RDF into encoding KIF as LISP-style 
Sexpressions, without violating any of RDF's own rules, in such a way 
that normal RDF turns out to be a valid special case of a KIF 
relational atom, ie [s V o] in RDF means (V s o) in KIF.  I thought I 
could see a simple extension which would do this, but I now see that 
all it enables me to do is to *describe* Sexpressions, not to 
implement them, and that the tool that I would need, in order to do 
what I wanted to do, is incompatible with the fundamental intuitions 
of RDF.

The RDF model, I now think, is really not properly captured by the 
connected blob-and-line diagrams which show connected graphs and 
suggest datastructures: it really is a *set* of isolated triples, 
which retain their meaning through any addition or deletion to the 
set; but those sets are not themselves in the model, so cannot be 
used as structuring tools. The graph-picture needs to be recomputed 
every time the set is changed.  If I have that right, it really isn't 
a base on which one can build any viable larger structures, seems to 
me. It has been made to resist any kind of stable extension being 
constructed. I feel like a mason who has been given teflon bricks.

Maybe we should stick to using RDF as a simple ground-data language, 
and just build or use something else altogether for doing more 
complicated stuff.

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 Friday, 8 June 2001 12:15:02 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:40 GMT