RE: AIN, NOBI and composition

Chris Ferris or IBM writes:
...

Where in the text that I have offered is the RFC2119 keyword "MUST NOT"?
The proposal that IBM has offered 
does NOT require that you know the unknowable (the complete set of
policy alternatives in the universe). 
The proposal we have offered is an attempt to make it clear that a
policy alternative is a complete 
expression of the set of behaviors to be engaged. 

If we have the "makes no claims" interpretation, then a policy author is
free to (for instance) exclude the 
security policy necessary to the interaction on the grounds that it is
too complicated. Thus, an endpoint 
wishing to interact with an endpoint whose policy was authored by this
lazy policy author would find 
out the hard way that the message needed to be signed and encrypted in
order to be processed. 

We believe that in order for policy to have value, it must be a complete
expression of the behaviors 
that are engaged for purposes of interaction. 

This has nothing to do with omniscience of policy assertion vocabularies
and exclusion of the 
set of behaviors implied by those absent from a given policy
alternative. 

IBM wants a policy alternative to be able to be taken at face value as
the expression of the 
set of behaviors that are to be used to interact (interoperate) with the
attached policy subject. 

Comment: Years ago, mathematicians used to say that numbers are the
things that make the Peano axioms true. They are just those things and
nothing else. Unfortunately logicians discovered model theory, and some
people named Lowenheim and Skolem noticed some interesting things. There
will always be other things (non standard models, they are called) that
make axioms true. The moral here is that in general trying to bound what
is in reality (behaviors) by saying that they are just those things that
make assertions true (that conform to the policy assertions) does not
really work-there will always be another set of behaviors (call them the
Oracle set) that will be a non-standard set of behaviors that conform to
the policy assertions the IBM set (your standard model) conforms to. I
think what can be said is

 

a policy alternative is able to be taken at face value as the expression
of _a_
set of behaviors that are to be used to interact with the attached
policy subject. (If the implementations are similar enough, they will
interoperate.)

 

It may be that the way one of these non-standard
implementations/behavior sets works happens to make them conform to some
other policy assertions not in the policy alternative. That is why we
should not be saying that that no other policy assertions are conformed
with except the ones in the policy alternative! It is enough that the
behaviors conform with the policy assertions. Full stop.

 

It seems to me that IBM is making some very restrictive assumptions
(policy assertions are each logically independent of one another, a
policy alternative is negation complete, the tacit logic is sentential
only, and that therefore reality is uniquely determined.) But even under
those highly restricted and implausibly strong assumptions the most you
will get is categoricity of models, and not identity of models (sets of
behaviors in this philosophy). 

 

So a policy alternative has a one to many relation with sets of
behaviors, a point that has been made earlier in this thread.