Re: PSEUDO-SCIENCE ?

[John Denker <jsd@MONMOUTH.COM>]

At 06:29 PM 2/25/00 -0500, Ludwik Kowalski wrote:
>
> When a teacher invests time to master a scientific area (such as geometry,
> algebra, electromagnetism or organic chemistry) s/he can be certain that
> what was mastered will always be useful in teaching the subject. But this
> is not true for those who invest in mastering software. A teacher investing
> a year or two in learning PASCAL or JAVA or WebCT may loose the mental
> investment when somebody decideds not to support the product.
>
> Something similar may happen in certain areas of engineering (for example
> a change from the tube-based to the semiconductor-based electronics) but
> it is never so drastic and so devastating as in the area of software. That
> is why I think that the so-called "computer science" is not really
> science. What is it?

At 04:35 PM 2/25/00 -0700, Daniel Schroeder wrote:
> Ludwik makes a good point, and makes it well.

Actually Ludwik made a number of points, some good points but also a few
far-from-good points.

I agree that there is too much "software rot" -- i.e. languages and
language features that go unsupported.  I hate it when I have a program
(many thousands of lines long) that compiled just fine a month ago but
won't compile now because some klutz "upgraded" the compiler.

On the other hand:

The subject line of this thread is horrifying.  If a topic is taught the
same way from year to year, it is evidence that it is a DEAD topic.  If a
topic changes from year to year, it is evidence that progress is being
made.  Calling it "pseudo-science" is pretty arrogant.

Also note that there is an ANSI standard for C, and an ISO standard for C++
and its powerful library -- so if you write lint-free code it should be
immune to rot for a good long time.

At 07:24 PM 2/25/00 -0500, Rick Tarara wrote:
> Well one does need to understand that 'Computer Science' is not just leering
> a current language.   People study data structures, information theory, and
> I'm sure lots of other stuff.


Lots of other stuff indeed.  Thanks, Rick.  The other stuff includes ideas
of timeless importance such as
 -- Shannon entropy, Shannon's coding theorem: a topic that several people
on this list have expressed interest in.  Surely this is one of the most
profound results of the 20th century.
 -- Turing's Halting Problem.
 -- Gödel's incompleteness theorem:  another of the most profound results
of the 20th century.
 -- Computational complexity, NP-completeness, and all that, with and
without quantum computation.
 -- Chaitin's algorithmic complexity.
 -- and lots more.

Calling this pseudo-science is not just arrogant, it's wrong.

A lot of so-called computer science courses are really computer
engineering, but still there's a big difference between engineering and
pseudo-science.  The label pseudo-science normally applies to things like
astrology.

There is also the notion of a "provably correct" program.  Advanced
computer science courses demand a level of formality and rigor that goes
far beyond what is seen in ordinary math courses.

Mathematicians argue over whether the Cooley-Tukey algorithm is "pure" or
"applied" mathematics.  Whether you call it math, science, or engineering,
it is a thing of beauty.

Rick continued:
>   And as Dan says,
> learning one computer language really does make it easy to move to another.

Right.

Where I work, a computer scientist who spent "a year or two" learning
Pascal or Java would be fired.  Somebody who took a week or two would be
embarrassed.  A day or two would be more like it.

=========

I've seen lots of bad software.  I've also seen lots of bad physics.  That
doesn't make it a pseudo-science.