Re: [RAF] small comment on discriminator "declarative vs. procedural"

The further discussion of this seems to have made Axel's point, but left his 
original issue in the dust.

Axel Polleres wrote:

> [In the current RAF wiki], I think that the discriminator
> "declarative vs. procedural"
> does not properly reflect the text:
> 
> " what was meant here is the availability (or not) in a rule of
>   procedures (expressed in some programming language) that perform
>   computations (such as doing arithmetic) that are difficult or
>  impossible to express in logic"
> at http://www.w3.org/2005/rules/wg/wiki/Rulesystem_Arrangement_Framework
> 
> Is IMHO misleading: In LP usually one understands the different between 
> procedural and  declarative in the sense that the order of rules does 
> not play a role for the semantics, i.e. that a set of rules is indeed 
> handled as a set with a declartive semantics. This is at least the main 
> issue which gives PROLOG its procedural style vs declarative logic 
> programming. It is not necessary about arithmetic or procedural 
> attachments alone.

I fully agree with this.  The distinction among explicitly ordered and 
implicitly ordered (stratified) and unordered (with Horn equivalence) rulesets 
is very much an issue.  The use of attached or built-in procedures in the 
antecedents and consequents and their relationship to the dynamic behaviors of 
the KB is a separate set of issues.  And among these latter issues is the 
relationship of attached procedures to rule ordering.

> So, I'd suggest to change the current heading for this discriminator
> to something like "builtins vs. no builtins" and redefine the 
> discriminator "declarative vs. procedural" as a new one.

I only care what the song IS, I am not all that concerned about what the song 
IS CALLED.  But Axel's point is that we need a discussion page for the 
semantics of orderings in rulesets, and that page is distinct from the 
semantics of interactions among functions, rule "execution" algorithms, and 
the KB.  And it is having two pages, not just one confused one, that I support.

As near as I can tell, Piero is on the rule ordering page, and Paul is on 
both.  (And this is no shock, because in commercial engines, explicit and 
implicit orderings are critical to the algorithms, and attached procedures are 
a sine qua non for having a customer base, and they interact in "wonderful 
ways".)

IMHO, we can't deal effectively with the "attached procedure" concepts (except 
with draconian limitations) until we have resolved the basic RIF capabilities 
and semantics with respect to "rule ordering" in a ruleset.

-Ed

-- 
Edward J. Barkmeyer                        Email: edbark@nist.gov
National Institute of Standards & Technology
Manufacturing Systems Integration Division
100 Bureau Drive, Stop 8263                Tel: +1 301-975-3528
Gaithersburg, MD 20899-8263                FAX: +1 301-975-4482

"The opinions expressed above do not reflect consensus of NIST,
  and have not been reviewed by any Government authority."

Received on Friday, 10 February 2006 17:12:13 UTC