Thoughts on Deterministic Processes and UMM

[David RR Webber <david@drrw.info>]

Team,

I sensed I was soap-boxing around deterministic processes and BPSS on 
Monday's call - so I
thought it worthwhile to take ten minutes here to type some brain-dumping.

A deterministic process actual has formal theoretical basis back in the 
days of Programming in Logic
and the development of Prolog - where there was work done to show that 
mathematically that the
decision matrix is a closed set.  Certain things are required for this - 
much of which will be very familiar
to BPSSrs - such as single entry point, single exit point (BTA), guard 
conditions that return true/false,
and then predicate clauses and fact assertions - aka variables, 
transactions and our BSI binding model.

Of course then we had the object oriented programming model emerge, 
along with object oriented Prolog
and then things like Java with the throw and catch method, and again 
Prolog contains constructs to
emulate those too.

The cool thing about a Prolog engine is that given a Prolog program it 
will actually evaluate it at parsing time
and tell you if it is deterministic or not.  That is obviously a very 
desirable behaviour from a BPSS / UMM
point of view.

So - looking at all this - I had already said earlier this year - that I 
believe a Prolog engine provides a
natural means to creating a validation service for a BPSS model.

Also - looking at where BPSS stands right now - I beleive I could take 
my VisualScript model - that
outputs a BPSS XML instance - and morph that - so that the exact same 
model representation can
output a valid Prolog program - with predicates, assertions and 
clauses.  Of course that is a non-trivial
exercise - so don't expect that in just a weekend - but the potential is 
there - to take something like the
excellent SWI Prolog (open source) foundation - and create such a BPSS 
validator.

I do beleive it is potentially very cool and important for adopters of 
BPSS to see that we are pursuing this
path of enabling deterministic process definitions when using BPSS.  Of 
course you *can* break the
model - hence Prolog will give you a warning when that happens - but the 
whole way we have set things
up to date positively reinforces and encourages deterministic models.

This IMHO is a good thing - and a thing of beauty.

OK - I'm indulged - you can go back to whereever you where ten minutes ago!

Thanks, DW