Re: [UCR] comment on reference to charter definition of "rule language"

On Mon, 6 Feb 2006, Peter F. Patel-Schneider wrote:
> I am puzzled by the following section of
> http://www.w3.org/2005/rules/wg/wiki/RIF_Use_Cases_and_Requirements
>
> 	1.1 What is a Rule Interchange Format And Why Create One?
>
> 	[...]
>
> 	A RIF is not a rule language.

I understand the RIF in general to be a set of languages with some common
features.  I usually think of this as a core language with a set of
extensions.  But I believe the two views are equivalent; every meaningful
combination of extensions can be given a name and called a new language.

The reason it's envisioned as a set of languages (instead of just one
language)  is that it appears we will need incompatible extensions.  As
the charter says:

  The general directions for extensions in expressive power lie along two
  roads: monotonic extensions towards full first-order logic (FOL) and
  non-monotonic extensions based on minimal-model semantics found in Logic
  Programming (LP) systems. [In phase 2] the Working Group will have to
  navigate this space and find extensions which best serve users.

This is also discussed in
http://www.w3.org/2005/rules/wg/wiki/Semantics_Extensibility_Point

There was discussion in that breakout and in the debrief which followed
which suggested this needs to be figured out in Phase 1 because the
different styles of semantics produce different results even for Horn
rules when queried via SPARQL.

Meanwhile, I think it's also true in a trivial way that any set of
languages which are self-identifying (as XML formats are) can be
considered as one language.  So in that sense at least, yes, RIF is a
language.

> To me, this opens up the distinct possibility of direct implementations of the
> RIF.

Surely it's impossible to tell -- from a sufficient distance -- whether a
system implements RIF logic directly or performs some translation to a
"native" rule language.

     -- sandro

Received on Tuesday, 7 February 2006 00:54:48 UTC