RE: [RIF-APS] Rules Sign

CSMA's proposal might not fit the current use of frames in PRD, but it
relies on the fact that almost ALL modern production systems make the
distinction between attributes of objects that refer to ONE ATOMIC value
-- such as one Person's age -- and attributes of objects that refer A
SET of values -- such as one Person's parents. In this case the rules
might propose some syntax to refer the set as a value and some syntax to
iterate across the elements of the set.

 

The current use of frames in PRD is biased toward a Prolog-style
interpretation. 

This is very nice as long as the Logic Programming community os
concerned, but I believe that is is a problem if we apply that to PRD
because for Production Rules systems, it neither fits the standard
practice nor the product offering for. 

 

Before entering this group I thought that standard proposals where
mostly about normalizing existing practice; this is not what is being
done here.

As this is my first participation to a standardization group and I
entered the group quite lately, I am happy to be educated on this, but
my intuition just lights a big red light. 

 

I am trying to share it with the team. 

 

 

 Patrick. 

 

 

-----Original Message-----
From: public-rif-wg-request@w3.org [mailto:public-rif-wg-request@w3.org]
On Behalf Of Michael Kifer
Sent: mardi 2 septembre 2008 17:40
To: Chris Welty
Cc: Jos de Bruijn; Adrian Paschke; public-rif-wg@w3.org
Subject: Re: [RIF-APS] Rules Sign

 

 

 

 

On Tue, 02 Sep 2008 10:44:37 -0400

Chris Welty <cawelty@gmail.com> wrote:

 

> Syntax for named arguments to use '('Name  TERM)')' instead of (Name
'->' TERM)

> Syntax for frames to use TERM '::' TERM instead of TERM '->' TERM

> Syntax for member to use TERM 'TY' TERM instead of TERM '#' TERM

> Syntax for subclass to us TERM 'SC' TERM instaed of TERM '##' TERM

 

 

You are proposing to replace perfectly good syntax with ugly
alternatives.

 

CSMA's proposal for using Name = Term is bad because it misleadingly
suggests

that there is only one value for Name, but in fact the value of Name is
a set

and Term is just one of the values in a set.

 

If you want to overhaul the syntax and free up -> for (classical)
implication,

then let's use something that mnemonically makes sense:

 

   a isa b

   c subclassOf cc or c sub cc

   name hasValue val  or name hasVal val

   etc.

 

We should use a different sign for rule implication both in BLD and in
PRD.

That should be => <= and not -> <- (provided that we agree on the
overall

overhaul).

 

 

      --michael