Re: RIF and QL

Dear Ed,

>Ah, but after 1650, French became the new "lingua franca" of Europe, and after 1800, it came even to be the dominant commercial language of the 
>Mediterranean.  So your re-interpretation of the term was once appropriate. 
>Only since 1900, with the rise of the British and American empires, has French 
>given way to English as the commercial lingua franca.  Il y a 150 ans, nous 
>serions obligés, selon la mode, discuter ce projet en français. ;-) 
>
:-)

>Et la 
>lingua franca prochaine peut etre chinois. :-o Mais, pour anglais, les grands 
>avantages sont CNN et l'Internet.  La technologie a changé aussi la 
>progression de l'histoire!
>  
>
Tu as rfaison.

I was aware that my translation of franca with French was not fully 
correct. I meant it as a joke. And remember, French tend to think, the 
French are the civilized Franks. I Franconia, part of the state of 
Bavaria I live in, this viewpoint is surpinsingly less widespread.


>
>  
>
>>>...
>>>      
>>>
>>One more personal opinion on this: I would like it much better if the 
>>RIF syntax would be pleasant to read and write -- at least the non-XML 
>>syntax. Eg, I would not be happy to have to write (a & B) => false 
>>instead of (logically equivalent) not (a & b).
>>    
>>
>
>I think the RIF standard syntax will be *only* XML.  
>
THat would be very sad. Look at XSLT. I believe that RIF needs a human 
readable syntax. Thjis is not much morte work and would considerably 
help RIF to achievre its goal, that I in RIF makes sense.

>I would say you have coupled precisely the two sides of the issue in this 
>sentence:  "sufficient expressibility" vs. "clear and simple semantics".  To 
>get clear and simple semantics you must give up some expressibility, or at 
>least you must make the expression of certain rules very complex.  So I think 
>there is a "tradeoff" here and the RIF WG must find a comfortable middle ground.
>  
>
I agree.

>  
>
>>>>Second, consider the following rules with non-monotonic negation:
>>>>
>>>>R2 = if not a then b
>>>>R3 = if not b then a
>>>>
>>>>The ruleset {R2, R3} might have been specified in a context where it is
>>>>used under the well-founded semantics. Under the well-founded semantics,
>>>>in the absence of other a- and b-headed rules, {R2, R3} means that a and
>>>>b are unknown.
>>>>
>>>>However, the ruleset {R2, R3} can also be used after the stable model
>>>>semantics where it means a or b is true.
>>>>        
>>>>
>>>This is somewhat inaccurate
>>>
>>>      
>>>
>>???
>>    
>>
>
>I said this because it is not quite "the absence of other a- and b-headed 
>rules".  Any rule with a or b in the consequent can affect the chaining in the 
>well-founded semantics case as well.  
>
I agree. I expressed myself improperly.

>And in the stable case, when you wrote 
>"a or b", you mean "a exclusive-or b".  
>
You are right, it is exor.

>But all of this is irrelevant to your 
>point, which is that the semantic model is critical to determining the valid 
>inferences from these rules.
>  
>
I agree.

>  
>
>>>>Third, consider a rule stating that a computer science student not 
>>>>attending the lecture on RIF must attend the lecture on Prolog.
>>>>...
>>>>        
>>>>
>>>We...ell, the rule is given in somewhat ambiguous natural language.
>>>What exactly is meant by "must attend"? 
>>>      
>>>
>>Forget the unecessary modality. Or remove it and take the rule as a 
>>database integrity constraint.
>>    
>>
>
>Well, actually that was the beauty of your example.  
>
It is conveyable without modal logic. Eg in DB integrity constraints. 
Arguably by relying on an assumption on the use of the formula, ie on 
what software/algorithm to use.

>I meant "an action" -- a specification for a human or automated activity.
>  
>
For me, "action" means a specification of a state transition, eg 
something changing some data.

>  
>
>>>So I don't deny the value of having the possibility of different model 
>>>theories.  I just want the ruleset to be explicit about the model 
>>>theory that would give it the intended interpretation.
>>>
>>>      
>>>
>>... and leave it to the user to change that model theory if he likes or 
>>needs.
>>    
>>
>
> From the point of view of interchange, both sides have to agree on what the 
>ruleset (of itself) "means".  That means I can send you a ruleset with the 
>rules above, and label it "stable-state", and you will interpret the rules 
>correctly and conclude a XOR b.  But you can integrate my rules into a ruleset with "well-defined" semantics, because you know what my rules mean and how to do that.  And you can use RIF to send your "well-defined" ruleset to someone else, but then you must label it "well-defined" and not "stable state".  In my view, the RIF must allow this -- it should not require all rulesets to have stable semantics, or all rulesets to have well-defined semantics.  
>
I agree.

>But it must also require me to say which semantics is intended for interpreting the ruleset.  (I suppose there could be a "default" semantics -- if you don't say, then it is safe to interpret the rules under model theory XYZ.)
>  
>
I agree. I would even add that several semantics (for different usages) 
might be specified for a same ruleset, with indication of different usages.

>But if, as a user, you choose to ignore the semantics I declared, e.g. because 
>your engine doesn't really internalize those semantics, then the inferences 
>you make from my rules may be "wrong" -- inconsistent with my knowledge.
>  
>
I agree. That often the (high!) price of communicating with others.

This remind me of this sad story of a French king "demanding" something 
to an English king -- meaning the French "demander" ("ask for") -- what 
led to a war. Maybe they all spoke French and the English misundertood 
"demander", I do not remember...

The way out is full standardization, ie the same rule language for all 
instrad of a RIF. (Or everyone speaking French, the secret dream of all 
French persons.)

> Well, we have time to think about this. This is the statement that 
> Enrico Franconi disagrees with. He believes that the "unifying 
> framework" for the several possible model theories is available 
> (through his work and the F-logic work) and that RIF can adopt a 
> unifying model theory and define a query language over that.

Let me call this the  "French approach", like in European history and 
recent EU politics: a full-fledge approach (proposed by the French), 
which all other must accept because it is the best approach (of course 
the best, it has been proposed by the French :-)). 

Opposite to the "French approach" is the "German approach": Agree on all 
it is possible to agree now and leave the rest for later. (With this 
approach we still have the German wondering if the Bavarian speak German 
and the Bavarian not asking themselves this question: They speak 
Bavarian, it is all what counts.)

The "French approach" sounds very good but does not work. The "German 
approach" works but is not the ultimate solution.
-- 
Francois

Received on Friday, 27 January 2006 11:13:44 UTC