Re: Data models as constraints

All,

We are working on a summary of today's telecon which we will post shortly.

Regarding this topic (data models as constraints):

It was my impression on today's telecon that most people present still do not 
understand the proposal, including Michael and Harold.  The latter two are critical, 
clearly, since they are editing the CORE spec.  In order to make progress on a CORE 
spec working draft released by the end of Feb, I am currently leaning towards having 
a 1st WD of the Core that does not include constraints, while continuing the 
discussion with the subsequent f2f meeting (probably in June) as a target for full 
resolution of the issue.  I note that from a practical standpoint, we still do not 
have a syntax and semantics for CORE constraints, and it seems unlikely they will 
materialize by the end of Feb.

I would like to resolve this formally at the next telecon, keeping in mind again that 
we are not proposing removing constraints from the core, just having a 1st WD without 
them.  The resolution woudl read something like:

PROPOSED: the CORE syntax and semantics in the first WD will not include constraints.

-Chris

Hassan Aït-Kaci wrote:
> 
> Gerd Wagner wrote:
> 
>>> Thanks for your comment on my paper: you must be the fastest reader
>>> that I have come across in my life! ;-)
>>
>>
>> In fact, I'm not a fast reader at all, of course, I just managed to
>> browse through your paper.
> 
> Now I understand better! :-)
> 
>>> Re. your comment, I must say thay I do not quite understand what it
>>> has to do with either the topic of the article, nor  what it means.
>>> What I wrote about is simple: because CLP makes rules and constraints
>>> orthogonal, it offers a way to use rules over arbitrary data models
>>> - not just data and FOTs. I then focused on the OSF and DL formalisms
>>> seen as constraint systems for objects and inheritance. The nature of
>>> the rules is *not* important and may be of all kinds including Horn,
>>> (with or without negations or all kinds), Production, YouNameIt, ...
>>
>>
>> OK, but you don't discuss that in your paper, do you?
> 
> Quoting from the paper (Section 3.3.2)
> 
> "Clearly, non-logical rules systems such as those based on
>  condition/action production rules typically used in experts systems
>  using object pattern-matching can also be expressed using OSF term
>  entailment (as opposed to unification).
> "
> 
>>>> How can you capture SQL, Prolog and production rules
>>>> without supporting NaF?
>>>
>>> Again the focus of my paper is *NOT* on the rules but (as the title
>>> states it clearly) about *data models as constraint systems*.
>>>
>>> Also, the nature of the paper is that of a (semi-formal) tutorial
>>> - not an exhaustive survey about all known or possible ways to
>>> conjugate rules and data. Perhaps you should write such an article?
>>> :-)
>>
>>
>> Come on, you are the expert here, not me!
> 
> You are as much as "expert" as I am I am afraid. My point though is
> that you seem to have a precise idea of what you want and you are the
> most competent to explain it.
> 
>> ...
>>
>>> At any rate, I am sorry that you find my work of debatable interest
>>> for your needs, and I apologize for disappointing your expectations.
>>
>> Your paper is clearly of high value as a publication, but for my
>> purposes I'd prefer to see a more RIF-focused version of it. 
> 
> Your personal purposes notwithstanding, one of the 2 reasons I wrote this
> paper and have shared it with the RIF/WG is that I was explicitly asked
> (by Jeff Pan, Alex Kozlenkov, and others at the last RIF F2F) to write
> such a document explaining in more formal details my claims that data
> models are constraint systems. The reason they requested it is because
> they could not see clearly how of CLP scheme proposal for the RIF that
> I have advocated without being sure how their own systems' data models
> could be seen as constraints. I accepted the challenge, agreeing that
> the onus was on me as the proposal's author to do so - hence the paper.
> The feeling shared by most RIfers (I believe) is that (1) it was not so
> obvious and (2) it was a topic of importance for the RIF. So it may be
> not "RIF"-focused, but I think that it is definitely "RIF-relevant".
> 
>> -Gerd
> 
> -hak

-- 
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 Tuesday, 16 January 2007 19:26:10 UTC