- From: Chris Welty <cawelty@gmail.com>
- Date: Mon, 14 Jul 2008 09:43:56 +0200
- To: Axel Polleres <axel.polleres@deri.org>
- CC: Jos de Bruijn <debruijn@inf.unibz.it>, kifer@cs.sunysb.edu, RIF <public-rif-wg@w3.org>
This appears to be unresolved. I see the problem Jos points out, that you can have the same symbol (and most problematically I guess if its the same URI) that *could* syntactically denote both an external predicate and an internal one. This *seems* to violate the design of the semantic web (URIs are rigid designators). </chair> Personally, I think the internal/external distinction is outside the scope of the semantic web anyway, so I don't think this matters. I am happy to leave it as is. It merely *allows* one to use the same symbol in different RIF contexts, it doesn't require it. <chair> Anyway, we need to wrap this up. Were you OK with things as they are, Jos? -Chris Axel Polleres wrote: > > Jos de Bruijn wrote: >> Michael, >> >> My argument is not about error-checking, but rather about the >> principle that the same constant you use in different places should >> mean the same thing. >> >> But I'm also willing to compromise for the case of external frames, so >> the compromise would be: >> >> - we create additional sets for external functions and predicates that >> are disjoint from the sets of "internal" symbols, and impose the >> condition that internal function and predicate symbols may not be used >> in external terms >> - we do not impose restrictions on symbols used in external frames >> >> Let's see what the rest of the working group thinks about this. > > Why do we need to make this restriction? It doesn't seem to buy us > anything, as semantically, the external funcs and preds are well > separated from the internal ones, so, no danger. > > Axel > > >> Best, Jos >> >> Michael Kifer wrote: >>> >>> On Thu, 10 Jul 2008 17:03:25 +0200 >>> Jos de Bruijn <debruijn@inf.unibz.it> wrote: >>> >>>> >>>> Michael Kifer wrote: >>>>> This was not an omission, but I am fine with separating external from >>>>> non-external symbols for functions and predicates. >>>> ok, good >>>> >>>> > As to the frames, I do not think any of the symbols >>>>> should be required to be external. >>>> But then the same constant used in different contexts has a >>>> different meaning, which I think was something we were trying to >>>> avoid in BLD. >>> >>> Frames are reflexive by nature. So, in some other statement you may >>> want to >>> say that some object (even external one) has a particular set of >>> properties and >>> list them). >>> >>> Frankly, I do not understand why is it a deal to allow the same, say >>> predicate, >>> to appear inside External(...) and outside of it. The reason for >>> separating the >>> symbols was to ease the interface with FOL. But separating external and >>> non-external symbols does not affect that. >>> >>> Syntactically it is clear whether a symbol is used as external or >>> internal, and >>> I see no reason to reinforce this with an additional syntactic kludge (I >>> conceded it just in the interests of peace :-). If your argument is >>> error-checking then it is not our business. Systems that care about >>> it would >>> build the appropriate error checkers. >>> >>> >>> --michael >>>> Best, Jos >>>> >>>>> >>>>> >>>>> --michael >>>>> On Thu, 10 Jul 2008 12:57:45 +0200 >>>>> Jos de Bruijn <debruijn@inf.unibz.it> wrote: >>>>> >>>>>> There is an issue in BLD, which I unfortunately did not catch >>>>>> before. I think it is probably an omission in the definition, but >>>>>> it is a substantive one. >>>>>> If we all agree that it is indeed an omission, we can probably >>>>>> address the problem, create a new frozen version, and vote about >>>>>> publication in the next phone conference on Tuesday. >>>>>> Personally, I am not ready to sign off on publication before this >>>>>> issue is resolved. >>>>>> >>>>>> The issue is the following: in the definition of well-formed >>>>>> terms, the set of all symbols is partitioned into predicate >>>>>> symbols, function symbols, etc. however, no distinction is made >>>>>> between external and "internal" symbols. The consequence is that >>>>>> the same function or predicate symbol can be used both in an >>>>>> external term and an internal term, and these two terms have >>>>>> different meanings, i.e., the same constant is interpreted >>>>>> differently based on the context, which is something we explicitly >>>>>> wanted to avoid in BLD. So, a built-in function may be used >>>>>> outside an external term and will be uninterpreted. >>>>>> >>>>>> The problem is easy to fix by defining additional sets of external >>>>>> predicate function symbols that are disjoint from the other sets >>>>>> of symbols and defining appropriate restrictions on external terms >>>>>> (i.e., the first function/predicate symbol in an external term >>>>>> must be an external symbol). >>>>>> It becomes a bit more tricky when considering external frames, but >>>>>> probably all constants used in an external frame should be >>>>>> external individuals/functions/predicates. >>>>>> >>>>>> Best, Jos >> > > -- Dr. Christopher A. Welty IBM Watson Research Center +1.914.784.7055 19 Skyline Dr. cawelty@gmail.com Hawthorne, NY 10532 http://www.research.ibm.com/people/w/welty
Received on Monday, 14 July 2008 07:44:44 UTC