Re: report on coverage issue

Dear Allen, dear All,


Two questions and two suggestions:

Questions:

1. Arn't we in the process of defining circular definitions like the
following: "A ruleset is covered by RIF if RIF can faithfully convey
that ruleset. RIF can faithfully convey a ruleset if this ruleset is
covered by RIF."? Why do we need the two notions "cover" and "faithfully
convey"? Why not simply choose one denomination and immediately give a
sensible definition for it, as suggested below?

2. Why to refer on the one hand to "rulesets", ie programs, on the other
hand to a language (RIF or RIF dialect)? This is inconsistent and could
be misleading, I am afraid.

Suggestions:

1. Why not simply define: "A RIF fragment conveys a rule language iff
this rule language can be translated in the RIF fragment in such a
manner that the (procedural and if available declarative) semantics of
the rule language is kept." The choice of the word "translated" is
conscious: this word has a wqell-established mewaning in Computer Science.

2. I would suggest to distinguish between language "fragments" and
"dialects". A fragment is a part of the language (like English without
the passive mode or Prolog without negation as failure). Dialects are
language (*)  that have sligthly different syntaxes and grammars but the
same expressive power as the lnaguages they are related to (eg Bavarian
or Swiss German are dialect of "High German", both with words that do
not exist in German and specific grammars. Another example are the
varios serializations of RDF: they are RDF dialects but, if they are not
incomplete, no fragments).

Regards,

Francois

(*) Please, let us avoid the old, and old fashion, discussion whether
(natural) dialects are (natural) languages or not. Modern linguistics
does not make such a normative, and in general polically-based,
distinction any longer.

Received on Tuesday, 19 September 2006 09:27:07 UTC