Layering (was Re: What do the ontologists want)

> Well, hey, you didnt mention layering until now. I suspect we are in 
> total agreement here, but what you call 'layering' is what I have 
> been calling 'implementation'.

Yeah!  :-)

> Model theory provides a much more flexible spec. Two people can use 
> completely different reasoners, as long as they both respect the 
> model theory (make only valid inferences). It gives you a lot more 
> flexibility and is much more secure against misunderstanding.
 
I need to do some more homework on this.


> > > 'conclusion'.  (You will also need to relate <a,b,c> to the three
> > > triples with 'RP' in the subject, but I presume that this wil be done
> > > by reification, so I won't dwell on it.)
> >
> >Reification of RDF statements of Layer 1 statements in a Layer 2
> >module, which you said this morning you had "NOOOOO problem with", I
> >believe.   Is it still okay with you?
> 
> I said I had no problem with saying that there was a triple and if 
> you believe it, time to forget it, ie treating a triple (for some 
> reason I keep typing 'trifle') as an assertion and saying something 
> about it. I didnt know about the layers thing, though, and that is 
> news. So is the RDF triple being trashed (ie disbelieved) in level 1 
> or 2 ? Im getting confused about where the assertions - the things 
> with the meanings, the things that are believed - are supposed to be.

Here's the text of that earlier example (mid:<200105172342.TAA01473@hawke.org>)

   <T, subject, A>
   <T, predicate, B>
   <T, object, C>
   <actionRequest, Forget, T>

The you-should-remove-it assertion was made in layer 2 about a layer 1
assertion.  The assertion to be removed was in layer 1. 

The interesting part is that you could remove the you-should-remove-it
assertion from layer 2 by removing the last of the above four triples
from layer 1 (with another four triples).

> Oh sure, I know it CAN be translated. But only if you know both 
> conventions. But look, isnt this the whole point of having something 
> like RDF, that it doesnt need all these ad-hoc translators, but 
> provides a kind of interlingua with a fixed meaning? If not, what the 
> hell is it for? We can write translators into and out of all kinds of 
> notations already (eg check out the Chimerea website and Stanford.)

I guess I'm more optimistic about the translators approach than the
interlingua approach, myself.  If one of the languages emerges from
the translation graph as the standard, then fine....   I hope the
languages can converge on some underlying structures, eg atomic ground
facts (with binary or n-ary relations -- nicely argued, btw).

> It has to know your intended meaning of those symbols. You can't 
> explain that meaning by writing RDF (the RDF semantics isnt up to 
> handling it.) So how is the engine going to know what to do? Unless 
> of course the bloke who wrote it had a manual for your notation that 
> you emailed to him.

Agreed.  As I understand it, until we have a sufficiently expressive
layer 2 module L which we can use to define the syntax and semantics
of another layer 2 module X, systems which read/write X will have to
be built by people reading manuals.  L itself will need to be defined
via manuals unless there is a sufficient L0 to define L, but I think
you're saying we can't build to more expressive languages this way,
and that matches my understanding.

   -- sandro

Received on Saturday, 19 May 2001 07:43:07 UTC