Re: UPDATE: status of longer version of layering document

>"Peter F. Patel-Schneider" wrote:
>
>>  The *next* step, of course, is to decide on which layering approach to use.
>>  I am delaying throwing out the ceremonial grenade for this next step until
>>  the current document has been at least somewhat assimilated.  However, if
>>  you can't wait, don't let my (unusual) reticence dissuade you.  :-)
>
>Discussion so far has focussed on how to salvage the first layering approach
>(same syntax, extended semantics), and rightly so in my opinion, 
>since this is the most natural option.
>
>
>1. The first fix to RDFS that would save the preferred layering 
>option has been dubbed "moving rdf:type into the meta-theory".
>Can someone clarify if this is the same as what was called 
>"stratifying RDF Schema" in earlier discussions? (ie not making 
>rdfs:Class an instance of itself). Stratifying the RDFS 
>class-hierarchy has been explored before by different authors (e.g. 
>[1],[2],[3])

Seems to me that this direction is a blind alley that would not 
achieve anything useful. Before going into this direction I would 
want to see a *very* clear explanation of why, and exactly how, 
making this change would solve any of the 'layering' problems. It's 
not enough to observe that it would kill one counterexample: you have 
to explain how it would avoid any similar paradoxical situation from 
arising.

This is a deeper issue than just layering on the SW, by the way. The 
current draft KIF semantics uses this same nonwellfounded 
construction (I stole it from there) and so if there are any deep 
problems with it, a whole other community needs to know about them as 
soon as possible.

>2. The second fix to RDFS that would save the first layering option 
>would be to distinguish "asserted RDF triples" from "not-asserted 
>RDF triples". The RDF(S) model-theory would only apply to asserted 
>triples. The triples that are "only OWL syntax" would be 
>not-asserted, would not contribute to the models, and would 
>therefore not construct paradoxes. Pat Hayes has been claiming that 
>RDF Core would seriously consider such an option.

I think they wouldn't actually *adopt* it, for essentially political 
reasons, as they are out of time; but they might be willing to 
recommend it for the immediate attention of the next RDF2 WG.

>It would not only solve our layering problem, but also solve the 
>problems of anybody else who tries to define other languages on top 
>of RDF. (For instance, McDermott has to resort to reification to 
>encode 'a->b' in RDF to avoid ending up with asserting 'a'. Allowing 
>unasserted triples would also solve his problem.)

Actually, McDermott has decided to just ignore this issue, which he 
considers to be not worth wasting time over. The system he is 
developing will examine any RDF graph and if it finds a parsable DAML 
'idiom' it will simply treat it as encoded DAML and ignore the RDF 
content. [McDermott, personal communication, DAML PI meeting] That is 
a robust 'ice-breaker' approach which, er, solves the problem by 
deciding to deliberately ignore it, and builds that decision into the 
software.

>
>3. I think it's a shortcoming of this (otherwise excellent) document 
>that it seems to suggest that the  same-syntax/extended-semantics is 
>not obtainable. It >*is*< obtainable with any of the above fixes to 
>RDF(S), and I would strongly suggest that the document discusses 
>them (after all, its purpose is to lay out the map of all options).
>
>Frank.
>    ----
>
>As a PS: I'm very happy that with the highly readable requirements 
>doc by Jeff et al. and the layering options doc by Peter and Dieter 
>this WG is well on its way to produce much more readable documents 
>than any of the other W3C groups I know of.
>
>[1] http://sesame.aidministrator.nl/doc/rdf-interpretation.html#N1c1e48d
>[2] http://www.semanticweb.org/SWWS/program/full/paper11.pdf
>[3] 
>http://www.kbs.uni-hannover.de/Arbeiten/Publikationen/2000/modeling2000/wolpers.pdf


-- 
---------------------------------------------------------------------
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, 15 February 2002 15:34:58 UTC