Path: utzoo!utgpu!watmath!clyde!att!rutgers!gatech!hubcap!steve
From: steve@ragman (Steve Stevenson)
Newsgroups: comp.lang.misc
Subject: Re: How to supplant FORTRAN (Was: Algol-68 down for the count)
Message-ID: <3762@hubcap.UUCP>
Date: 6 Dec 88 13:25:40 GMT
References: <814@quintus.UUCP>
Sender: news@hubcap.UUCP
Reply-To: steve@ragman
Lines: 48

From article <814@quintus.UUCP>, by ok@quintus.uucp (Richard A. O'Keefe):
> This sounds important, but I don't understand what you mean by
> "a derivation criterion".  (I guess Statistics doesn't count as science
> or engineering, otherwise I would know what the "recognised ways" are.)
> Where does program transformation come into this?

What I mean is to take a physics model and postulate some behavior.  That
behavior is subject to principles which have their effect on the
trajectory functions.  Take for example the Navier-Stokes equations
for fluids.  It specifies the various parameters, etc.  The development
proceeds until either an analytic solution exists or we must
solve numerically.  The bridge is then to Numerical, which must
deal with the convergence, etc.  There is the ultimate test of the
postulates: does the model validate.

Statistics does qualify in this sense, because we must provide sampling
information --- and a bridge to probability.  Again, the questions and
parameters are well understood, or the statistician must then deal
with a ``first principles'' problem.

The history of science is replete with
stories of models not being accepted, even though they eventually were
based on empirical evidence.  Example: statistical mechanics and its
sad (personal) history of carpings and suicide.  I know several folks
who were burned bad by the computer in the late 60's.  The program was
fine: the model stunk and folks forgot to validate.

[This goes back to the discussion I started on sci.logic.  We have a
real handle on the (formal) Reals.  One is unlikely to propose a model
of first order logic without producing a deduction theorem, for example.
That discussion is aimed at the (pragmatic) observation that
a ``proof'' isn't a proof if no one believes it.  A derivation is not
a derivation if no one believes it or can validate it.  Now, the
sum total of the ``mores'' of the community is the ``philosophy'' of
the subject.  Who set the p<.05 mark for most tests --- the community
did.  Can we have the same derivation concept in CS?]

I think that the use of language between scientists and computer scientists
is quite different: the scientist seeks insight --- that's the product.
Yours and my products are meant to be a model of thought for the next
person.  We tend not to directly provide insights, but let the user
come to her/his own.  I'm proposing that we look at the ultimate
objective: providing the scientist with insights; that subsumes our
original goal of providing models of thought --- albeit only indirectly.

Steve (really "D. E.") Stevenson           steve@hubcap.clemson.edu
Department of Computer Science,            (803)656-5880.mabell
Clemson University, Clemson, SC 29634-1906