[RIF] Reaction to the proposal by Boley, Kifer et al

Hi,

This is a statement stronlgy in favour of the proposal by Harold Boley,
Michael Kifer, Jeff Pan, Gerd Wagner, Alex Kozlenkov, Jos de Bruijn,
Mike Dean and Giorgos Stamou to first define a *language of conditions*
to be shared by the RIF dialects.

A language of conditions is clearly needed for all use cases and all
dialects that have been mentioned so far. Specifying a language of
conditions would be a first important step in the WG's activity. The
syntax and semantics proposed are good choices, probably the most
reasonable choices that can be made.

The following amendments to the proposals are suggested for pragmatic
reasons, i.e. for easing applications. These amendments do not
significantly affect the proposal, especially they do not significantly
affect its Semantics part.

1. Syntax

Connectives 'And' and 'Or':

In addition to binary 'And' and 'Or' connectives, we suggest n-ary 'And'
and 'Or' for all n (either with n >= 1 or else with 0-ary 'And' meaning
'Top', i.e. a formula true in all interpretations and 0-ary 'Or maning
'Bottom', i.e. a formula false in all interpretations).

Connectives 'Xor':

A binary, or preferably n-ary 'Xor' for all n. (Remark: 'Xor' amount to
negation. Hence 'Xor' should come in two flavours, a 'Neg' and a 'Naf'
flavours).

Extended existential quantifiers:

Exists at least n (with n>=1).
Exists at most n (witrh n >=1).
Exists [n, m] (with m>=n>=1).

(Remark: Exists at most n amounts to negation, hence two flavours of it
should be considered.)

Types for objects should be considered, too.

Other pre-defined relations on data or objects:

If a type has an equivaelnce or order relation (e.g. < on integers),
then this relation should have the same status in the 'languiage of
conditions' as = on this type.

'Neg and 'Naf' negations:

Syntax making it easy to turn a 'Neg' negation into a Naf' negation and
conversely in conditions.

2. Semantics:

Types as defined in RDFS and a RDF application should preferably be part
of the Semantics of the 'language of conditions'. This would make it
possible firstly to define types on objects, secondly not to have to
re-express RDF(S) types in RIF.

(Remark: The only difficulty we see there is that RDF(S) not precluding
cyclic sbclass relationships, this could lead to barber paradoxes in
RIF. A simple solution would  simply be to only accept RDF(S) types if
they do not express cyclic sublcass relationships.)


Regards,
François and Paula