General talk about software patterns

[Dan Palanza <dan AT capecod.net>]

Hi Mike,

> > I found this statement both
> enlightening and amusing:
> > "'In 20 years time perhaps all computer systems 
> > will be built on a theory that is understood. 
> > We are trying to establish these theories.' "
>
> The quote as I understood it was only related to
> one of the projects:
>
>         2. Science
> for Global Ubiquitous Computing 
>
> > But "theory" implies a mathematical solution. 
>
> I disagree.  Theories are explanations sometimes
> backed up with empirical data.  They can be mathematical
> or not.

I do understand that "theory" is widely used in
non-mathematical contexts. I am proposing that the misuse of the word is
a major problem when it comes to issues of social science. If you don't
mind me bringing in bookkeeping for a moment, which I am increasingly
seeing as a social science problem. When I speak to mathematicians about
double entry bookkeeping, sometimes consciously, and sometimes
sub-consciously, they seem to me to be trying to get my words to fit a
mental model they have that if what I am saying is to prove to be
correct, at some point, the image of my argument must fit a mathematical
notion of theoretical proof. With that mind set on their part, even the
simplest things I say fall on deaf ears.

In fact, my research is increasingly validating a fact that the rationale
for bookkeeping's solution to social science issues have a reciprocal
rationale to the mathematical theorem. Where in mathematics one begins
with the logical assumption of an axiom, which gets tested in empirical
proofs, which in turn lets the assumed axiomatic model serve as a
functional calculation of future behavior of the system, double-entry
bookkeeping, as a pattern language, does just the reverse.

In bookkeeping, as pattern language, what is assumed is not a fixed
physical axiom, it is a variable intellectual container--an account that
receives an accounting. As in mathematics there is still a relation
between the physical system of facts being interpreted through
measurement and an intellectual language of accounts being compiled
through _expression_. But where axioms are focusing on rectilineal and
rectilineal rotational forces, as the behavior of the system being
studied, bookkeeping is focusing on the identification of informational
fields subject to signals that change the information within such
fields.

Bookkeeping does measure as fact the value of a trade, but it makes no
predictions based upon such facts. It uses those values to validate
future states of the system in two ways: Profit [loss] and Balances that
reflect future potential trades.

> > Is there a single
> new "theory" that mathematics has 
> > added in the past 40 years for the benefit of 
> > understanding computing? 
>
> I would say yes, for example Grenander's General Pattern
> Theory.

I have not studied it, but I would be willing to bet that if it does form
a pattern language, then it will encounter the same set of issues that
bookkeeping encounters.

> Its applicability
> extends _all_ domains from visual
> pattern recognition, language, medical, software ... 
> you name it.
>
> But there are many others, of course.

Yes, I have read a number of contemporary math books that increasingly
work logic into mathematics. But just doing that without defining the
implications of two reciprocal rationales--one lineal and one
nonlineal--raises questions. My bet is that in all cases you will find a
clear separation between the traditional mathematical interpretation of
physical behavior as the system versus a non-traditional compilation of
intellectual identification as the system's control language. Calling the
non-traditional compilation of intellectual identification as the
system's control language "mathematics," I find, serves no
useful purpose.

> > It seems to me
> that if progress toward understanding 
> > computer systems will be made in the next twenty years 
> > that progress will begin when a community of users 
> > gets their fossilized mathematical reasoning about 
> > "theories" out of the way. 
>
> Well, I agree with this.  But I think including
> biomimetic or even biological theories (that are
> not necessarily mathematical), is a good start.

I would rather argue that they are pattern languages, for the simple
reason that they involve decision control behavior within the system of
study itself, which does not happen in traditional mathematics. In
traditional mathematics, once the theorem is proven the pattern is
invariant. In biology we must do an accounting because the patterns are
variable due to decision control within the system of study itself.
Because the patterns are variable we call their accounting "pattern
language." Even if all the individuals at birth inherit one same
template of rules, their behavior, in patten form, may vary
widely.

> > The computer is a
> social science problem. Social 
> > science must deal with decision control. 
>
> I don't know if the "computer" is a social science
> problem, but I would agree that Software Development 
> is mostly a social problem. 
>
> > What mathematical theorem will withstand the need 
> > for a proof that tests for how you or I will decide 
> > to behave tomorrow? Or, for that matter, to test 
> > for what laws you or I might decide to enforce on 
> > other people's behavior?
>
> Localized Nash Equilibriums constrained by common laws?
>
> You can probably call this an imposed "moral imperative"
> :-)

Mike, when Western science came into being it pushed out a widely
supported art form called "Black Magic." Science didn't
disprove Black Magic, is simply proved to be so magical in its own right
that fewer and fewer people were willing to support half-baked
competitors like Black Magic.

When the present economic community comes to understand double entry
bookkeeping, what passes for economic theory in our times will go the way
of "Black Magic;" that's a promise you can count on.

Dan