Re: [RIF] Extensible Design

If you buy the primary goals of the RIF (to exchange rules, in a  
widely deployable way), then you will *not* accept the multiple  
languages approach.

It is way too complicated. And it strikes me as a cop-out (to use an  
Englishism).

There is a balance to be struck between enabling exchange and  
respecting the original. It does not mean that that balance has to be  
on the respecting end of the spectrum.

An alternative approach is to focus on the kinds of rule information  
that needs to be exchanged (leaving on one side for the moment the  
question of semantics) and then doing so in the simplest possible way.

I strongly suspect that that will not be possible without compromise;  
our task is to find that fulcrum.

For example, I suspect that there would be very little argument on  
the need for exchanging ground facts. However, it is not obvious to  
me that the RIF needs to support RDF to the extent of allowing RDF  
triples as the consequent of a rule (just to pick a random example).  
The reason: we are supposed to focus on the exchange of rules, not  
the generation of new inferences. Nor is it our mission to 'fix' RDF.

I think that we also need to exchange rules with variables in them --  
especially when the exchange is between tools rather than across the  
Internet.

But note, that already, rules with variables strike many people as  
very complicated. (Another puzzle: distinguish rule exchange from  
virus propagation).

I think that somebody who took a principled stand to stubbornly  
refuse all features until each one was proved to be essential could  
be of great service to this group. (Not me!)

Frank







On May 4, 2006, at 1:45 PM, Peter F. Patel-Schneider wrote:

>
> From: "Gerd Wagner" <wagnerg@tu-cottbus.de>
> Subject: RE: [RIF] Extensible Design
> Date: Thu, 4 May 2006 22:31:55 +0200
>
>>>> Hopefully there can be many modules shared between
>>>> dialects, where both the syntax and semantics are
>>>> shared. I'm not sure if it'll ever make sense to share
>>>> syntax but not semantics for some part of a language.
>>>
>>> Well, it seems to me that the proposal by Boley et al
>>> advocates precisely this view. My reading of the
>>> proposal is that several (perhaps many) RIF dialects
>>> will share the same syntax (or very similar syntaxes)
>>> for conditions but will diverge on semantics.
>>
>> Let me try to elaborate on this observation:
>>
>> 1) The RIF family will consist of several branches
>> of dialects, most of which overlap in their condition
>> language. Each branch will have a core, which
>> defines the common syntax and semantics of the branch.
>> Extensions of this core define additional syntax and
>> semantics.
>>
>> 2) Most RIF dialects will not only share the syntax

>> but also the semantics of conditions (except for
>> normative/integrity rules, which do, in general, not
>> have conditions).
>
> Hmm.  I don't read this in the proposal at all.  My understanding  
> is that
> in the proposal the semantics of conditions varies between FOL  
> dialects and
> LP dialects and even varies between different LP dialects.
>
>> 3) Data literals, object names, function symbols
>> and predicate symbols may be typed. Using suitable
>> predicate/atom types, this allows to represent RDF
>> and OWL rules directly (and not only via a "query
>> interface").
>
> Again, I don't see this in the proposal.  (Not that I don't think  
> that it
> is a good idea, however.)
>
>> -Gerd
>
> peter
>

Received on Friday, 5 May 2006 14:39:27 UTC