[
Lists Home |
Date Index |
Thread Index
]
The main value of simulation is the capability
to evaluate competing hypotheses without incurring losses
of assets (the danger of using real time orchestration over
production assets for some kinds of event response). I
think we agree there.
Returning to the feedback model posed earlier, the abduction
phase in which one chooses the subjects of interest is hard
to simulate. Choosing inadequate datasets is a primary
cause of failed simulations. Knowing when one has enough or
the right data is hard to do. Waiting to get more data can
be deadly, and sometimes, more data doesn't improve the
analysis. Ask yourself, regards XML Schema design, when
do I have enough representative document instances to
accurately model a class of documents?
Things only get a little better in the
second phase of induction, but it is precisely this phase
that spawns multiple hypotheses. Simulation can be of
great benefit to the inductive phase as long as the situation
analysis produced accurate/timely models and the corresponding
theoretical models account for all of the observables.
In short: competing multiple hypotheses are a better means.
Simulations should enable comparisons of real world measurements
against the predicted results. If you think about the ways in
which we design schemas, there are obvious correspondences
to gathering multiple documents and feeding them to the validator
while constraining and loosening the type definitions. Validation
error logs should be informative to an analysis, not just an
audit trail.
Also, just because the simulation shows it can work doesn't mean
it will. The installed base has to be in place first. So the
market cycles of individual integrated technologies must
be factored into the analysis. Again, Chertoff quoting from
the NIMS/NRP on CNN was revealing. Another problem is
beggaring a technology by specification yet assuming it is
a standard for actualized technologies. GJXML is a stunning
example of attempting to beggar a technical implementation
over understanding the installed base. In both cases, a
wishlist for outcomes became more important than the real
time actualized components. We have no problems committing
to achieve the goals of these, but making real time decisions
based on them is wishful thinking at its worst.
len
|