Conference, Modeling

CSGnet Archive CSG2002
1

[From Rick Marken (2002.07.01.0900)]

First, I would like to encourage all PCT enthusiasts who can possibly
attend the CSG conference to do so. The conference is a definite go and
it will be held from July 24 - 28 at Aqueduct Center in Chapel Hill, NC.

Next, I would like to see if we can start a discussion of modeling, both
for now and during the conference. This interest is motivated by an
article that Chuck Tucker recently sent me. The article, by Lissack and
Richardson (who, I presume, are sociologists), is called _When Modeling
Social Systems, Models <> The Modeled_ . It is a "reaction" to a book by
Wolfram called _A New Kind of Science_. Apparently Wolfram thinks that
you can learn all about living systems by just looking at how models
behave. Lissack and Richardson worry that this may stop the funding of
observational research. I've ordered Wolfram's book so I don't know what
point he is really trying to make. But Lissack and Richardson don't
seem to get what modeling is about (and if their description of
Wolfram's book is accurate, Wolfram doesn't see to get it either), at
least not from my point of view. They think modeling is, at best, to
help conceptualize underlying processes. This is partially true. But I
see models as an attempt to represent the structural and functional
characteristics of a system that are responsible for the behavior we
observe. Models are a guess at the nature of the "reality" responsible
for what we perceive. They are an essential component of the scientific
process.

The Lissack/Richardson article as well as some other off-line
discussions have suggested to me that the PCT approach to modeling --
which I think is equivalent to the approach to modeling in the
"physical" sciences" -- is not really part of mainstream
behavioral/social science thinking. I think this may be one of the main
reasons for the lack of acceptance of PCT. I realized that most of the
lack of acceptance of PCT takes the form of _acceptance_ of PCT. Many
social scientists agree with PCT as it is verbally described: they
reject cause-effect thinking, they agree that behavior occurs in a
closed loop, they think that perception is a central component of
behavior, etc. This is why we so often hear that PCT is nothing new, or
"nothing but..". So PCT has really suffered, I think, not from rejection
but from _acceptance_. What's accepted is verbal descriptions of PCT,
which can be made to mean what the listener wants to hear.

This, I think, is why modeling is essential. People who do modeling and
understand what modeling is about are not likely to think that PCT is
"nothing but..." something else -- because they can see that the PCT
model doesn't function like that something else. Unfortunately, PCT is
often "up against" other verbal theories rather than other models. This
makes things even more difficult.

So maybe the way to get PCT across is to explain what modeling is
_about_ from the PCT perspective rather than describe what the PCT model
_is_. I think that most of the theories we rail against --
cause-effect, caused output, reinforcement, manual control, etc -- can
be rejected, not as being incorrect theories but as being _non-models_
in the scientific sense.

Best regards

Rick

···

---
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

2

[From Bruce Nevin (2002.07.01 13:40EDT)]

Rick Marken (2002.07.01.0900) --

···

At 09:00 AM 7/1/2002 -0500, Richard Marken wrote:

[...] the PCT approach to modeling [...] is not really part of mainstream
behavioral/social science thinking. I think this may be one of the main
reasons for the lack of acceptance of PCT.

A very astute summation. Thanks.

I hope we have a good set of tutorials in modelling at the conference.

         /Bruce

3

[From Rick Marken (2002.07.01.1400)]

Bill Powers (2002.07.01. 1326 MDT) --

For some time, at the urging of David Goldstein, I've been trying to think
of ways to approach modeling without a computer or mathematics. It's
basically impossible to go all the way, but there are certain aspects of
modeling that can be put into practice almost anywhere and by almost anybody.

I think this is an important idea. Any discussion of modeling, especially at the
CSG meeting, should be accessible to people who are not modelers. I believe that
the vast majority of people at the CSG meeting will have a practical rather than
a scientific interest in PCT. Therefore, I don't think workshops on alternative
ways to implement models on a computer (using Delphi, Vensim, etc) are going to
be of much interest to most of the people attending the conference. What I was
hoping for was something like what you describe above: a way to describe
modeling in general (and the PCT model in particular) sans mathematics or
computers. I think that would be a great contribution for everyone involved in
PCT -- those who with practical _as well as_ those with a scientific interest in
PCT (and, of course, to those, like me, who are interested in both).

Best regards

Rick

···

---
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

4

[From Rick Marken (2002.07.01.1430)]

Stephen O'Shaughnessy wrote:

For my 2 cents I think a computer language used for modeling should be
object oriented. I vote for Java. It's inexpensive (Free unless you want a
nice IDE) and runs on many platforms. I've never learned Pascal, is Delphi
OO? I've never learned Smalltalk either, but I would be willing.

I second your vote for Java! It's not the OO aspect that matters to me. It's
the "_free_ that matters to me. As does the fact that it's cross platform,
easily accessible to everyone and easily distributed on the web.

There are things about Java that are difficult (and annoying) for an old
fashioned procedural programmer, like me). But these can be easily worked
around. If the goal is to allow a network of highly dispersed people to develop
and share models of behavioral phenomena, then Java is the way to go.

I think what we need is someone who can teach the basics of Java as clearly and
simply as possible, with an emphasis on using Java for modeling and, perhaps,
interactive experiments (like mine on the net). I'm not the one who can do
that, though. I struggle along with Java as best as I can. But maybe you are?

Best

Rick

···

--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

5

[From Bill Powers (2002.07.01. 1326 MDT)]

Rick Marken (2002.07.01.0900)
Bruce Nevin (2002.07.01 13:40EDT)

I hope we have a good set of tutorials in modelling at the conference.

        Bruce

The modeling theme is a good one for the conference. Considering how little
we've done in the recent past along these lines, however, I think that we
need some workshops first to figure out exactly what we want to do. Dick
Robertson is no doubt frustrated by the lack of action concerning his
proposals, and Wolfgang Zocher must be even more frustrated by the illness
that has prevented him from working on the graphic version of SimPCT.
Christian Heider (CSG Germany) proposed that we adopt Smalltalk as a
simulation language, and of course through Mark Abrams we became aware
of Vensim, which some of us have tried out. Now we're looking into
Delphi, which is really Pascal programming for Windows.

The most straightforward approach (easy programming, minimum of ideology
built in, ability to pass models around via internet) is Zocher's
SimPCT. Unfortunately I no longer can put my hands on the source code, so
we would either have to reconstruct it (I actually wrote the first very
simple version of it, so could probably do it again), or we would need to
start with basic specs and build our own system from scratch. Ideally it
would work on Macs as well as PCs, which may point us toward the free
versions of Vensim, which I think have a Mac version. These are some of the
things we need to talk about in a workshop.

Anyway, that's only part of the deal, the part that lets everyone try
models on their own machines and pass around models they've designed. We've
done some of this before, with mixed results. Perhaps just as important is
to discuss modeling itself -- how to set up a generative model of
real-world phenomenon so its behavior actually means something. The
computer program stuff is concerned with implementing models at the logic
level, but to organize that we have to think about the principles of
modeling, too. I don't know of any texts on this subject that are less than
about 50 years old, so maybe we can meet with the purpose of generating at
least an outline of a small text, in the form of principles that can be
taught through examples.

For some time, at the urging of David Goldstein, I've been trying to think
of ways to approach modeling without a computer or mathematics. It's
basically impossible to go all the way, but there are certain aspects of
modeling that can be put into practice almost anywhere and by almost anybody.

I don't have this worked out coherently, but the general idea is that a
model is an explanation of phenomena in terms of underlying processes, and
when you propose a model you're also committing yourself to certain
_predictions_. The structure of a true model is such that under various
circumstances it will act according to its properties to produce specific
consequences. If you have stated the structure of the model completely
enough to have a model at all, then anyone can propose conditions or
circumstances to which the model MUST respond in way that can be worked out
from its properties. If you then set up those same conditions or
circumstances for the real system, the model's behavior constitutes a
prediction of the real system's behavior. This is the ultimate test of the
model's correctness.

Doing this on a computer is one obvious application of this idea, but what
if you can't come up with that sort of model? Is it still possible to give
an explanation of behavior in terms of underlying properties, and then to
use those properties to develop a reasonably rigorous prediction of how the
model would behave under novel conditions? I feel that the answer is yes,
but right now that's as far as I can go. I think this would be worth
pursuing at the meeting -- another facet of the modeling theme.

I'm going to this meeting just about empty-handed (and -headed), but if
we're all thinking about this theme between now and then I believe
something will jell (gel?).

Best,

Bill P.

6

from Rick Marken (2002.07.01.1530)]

Acree, Michael" wrote:

Can you give us the reference for the Lissack and Richardson article?

I wish I could. The paper Chuck sent to me gives no indication of where (or whether) it was published. I think Chuck is trying to get that information for me now. Maybe he could put it on the net himself. OK Chick?

Best

Rick

···

--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

7

For my 2 cents I think a computer language used for modeling should be
object oriented. I vote for Java. It's inexpensive (Free unless you want a
nice IDE) and runs on many platforms. I've never learned Pascal, is Delphi
OO? I've never learned Smalltalk either, but I would be willing.

Steve O

···

-----Original Message-----
From: Bill Powers [mailto:powers_w@FRONTIER.NET]
Sent: Monday, July 01, 2002 3:16 PM
To: CSGNET@listserv.uiuc.edu
Subject: Re: Conference, Modeling

[From Bill Powers (2002.07.01. 1326 MDT)]

Rick Marken (2002.07.01.0900)
Bruce Nevin (2002.07.01 13:40EDT)

I hope we have a good set of tutorials in modelling at the conference.

        Bruce

The modeling theme is a good one for the conference. Considering how little
we've done in the recent past along these lines, however, I think that we
need some workshops first to figure out exactly what we want to do. Dick
Robertson is no doubt frustrated by the lack of action concerning his
proposals, and Wolfgang Zocher must be even more frustrated by the illness
that has prevented him from working on the graphic version of SimPCT.
Christian Heider (CSG Germany) proposed that we adopt Smalltalk as a
simulation language, and of course through Mark Abrams we became aware
of Vensim, which some of us have tried out. Now we're looking into
Delphi, which is really Pascal programming for Windows.

The most straightforward approach (easy programming, minimum of ideology
built in, ability to pass models around via internet) is Zocher's
SimPCT. Unfortunately I no longer can put my hands on the source code, so
we would either have to reconstruct it (I actually wrote the first very
simple version of it, so could probably do it again), or we would need to
start with basic specs and build our own system from scratch. Ideally it
would work on Macs as well as PCs, which may point us toward the free
versions of Vensim, which I think have a Mac version. These are some of the
things we need to talk about in a workshop.

8

[
From Bill Powers (2002.07.01.1537 MDT)]

  Stephen O'Shaughnessy (2002.07.01) --

For my 2 cents I think a computer language used for modeling should be
object oriented. I vote for Java. It's inexpensive (Free unless you want a
nice IDE) and runs on many platforms. I've never learned Pascal, is
Delphi OO? I've never learned Smalltalk either, but I would be willing.

I think that's one level down from the modeling language needed for most
non-programmers on this net. Vensim is a high-level language in which you
can draw functional blocks and interconnect them, then run the simulation
and plot the results. SimPCT was similar except for the graphics, which
were under development. The idea is to simulate an analog computer on a
digital machine.

For my own part I much prefer procedural languages like C or Pascal --
somehow I''ve never been able to get the rationale of OOP into my head. I
think that in part I'm resisting the idea that an object can do things to
itself. OOP strikes me as ideological. Also, writing event-driven programs
.is a very awkward way to simulate control systems, especially if real-time
interactions are also needed. But that's Windows' fault, I guess, not OOP's.

Delphi is object-oriented for sure, but fortunately for me it's possible to
use the "Forms" without really understanding them, and write code that's
almost purely procedural. If you like OOP, Delphi may be the way to go,
since we can both write programs that will run, but each can do it in the
most comfortable way.

Anyway, whatever the underlying source code looks like, the users of our
simulation product will not see that level of detail.

Best,

Bill P.

9

[From Rick Marken (2002.07.01.1710)]

Bruce Nevin wrote:

Mike Acree (2002.07.01.1521 PDT)--

>Can you give us the reference for the Lissack and Richardson article?

A google search turns up this:

Emergence, 19 Jun 2002, pp. 1-13, "When Modeling Social Systems, Models
[Not Equal] The Modeled," Michael R. Lissack and Kurt A. Richardson.

Of course! The Web! Nice work!

It turns out that _Emergence_ is an electronic "Complexity" oriented journal (www.emergence.org). So maybe after the meeting we can submit the results of our
deliberations on Modeling for publication in _Emergence_.

Best

Rick

···

--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

10

[From Mike Acree (2002.07.01.1521 PDT)]

Rick Marken (2002.07.01.0900)--

The article, by Lissack and
Richardson (who, I presume, are sociologists), is called _When Modeling
Social Systems, Models <> The Modeled_ . It is a "reaction" to a book by
Wolfram called _A New Kind of Science_. Apparently Wolfram thinks that
you can learn all about living systems by just looking at how models
behave.

Can you give us the reference for the Lissack and Richardson article? My reading of Wolfram is that he thinks the universe _is_ a cellular automaton (though a somewhat more complicated one than those he has studied), with both space and time being discrete.

Mike

11

A google search turns up this:

Emergence, 19 Jun 2002, pp. 1-13, "When Modeling Social Systems, Models [Not Equal] The Modeled," Michael R. Lissack and Kurt A. Richardson.

The original page at www.wolframscience.com/media.html displays blank for me, but Google's cached page is at

http://216.239.51.100/search?q=cache:QsL7sRroZukC:www.wolframscience.com/media.html+Lissack+Richardson+When+Modeling+Social+Systems+Models+Modeled&hl=en&ie=UTF-8

···

At 03:24 PM 7/1/2002 -0700, Acree, Michael wrote:

[From Mike Acree (2002.07.01.1521 PDT)]

Rick Marken (2002.07.01.0900)--

>The article, by Lissack and
>Richardson (who, I presume, are sociologists), is called _When Modeling
>Social Systems, Models <> The Modeled_ . It is a "reaction" to a book by
>Wolfram called _A New Kind of Science_. Apparently Wolfram thinks that
>you can learn all about living systems by just looking at how models
>behave.

Can you give us the reference for the Lissack and Richardson article?

12

{From Samuel Saunders (2002.07.01.1645 MDT)}

Bill Powers (2002.07.01.1537 MDT)

  Stephen O'Shaughnessy (2002.07.01) --

>For my 2 cents I think a computer language used for modeling should be
>object oriented. I vote for Java. It's inexpensive (Free unless you want a
>nice IDE) and runs on many platforms. I've never learned Pascal, is
>Delphi OO? I've never learned Smalltalk either, but I would be willing.

I think that's one level down from the modeling language needed for most
non-programmers on this net. Vensim is a high-level language in which you
can draw functional blocks and interconnect them, then run the simulation
and plot the results. SimPCT was similar except for the graphics, which
were under development. The idea is to simulate an analog computer on a
digital machine.

For my own part I much prefer procedural languages like C or Pascal --
somehow I''ve never been able to get the rationale of OOP into my head. I
think that in part I'm resisting the idea that an object can do things to
itself. OOP strikes me as ideological. Also, writing event-driven programs
.is a very awkward way to simulate control systems, especially if real-time
interactions are also needed. But that's Windows' fault, I guess, not OOP's.

Delphi is object-oriented for sure, but fortunately for me it's possible to
use the "Forms" without really understanding them, and write code that's
almost purely procedural. If you like OOP, Delphi may be the way to go,
since we can both write programs that will run, but each can do it in the
most comfortable way.

Anyway, whatever the underlying source code looks like, the users of our
simulation product will not see that level of detail.

Best,

Bill P.

While we are listing possibilities, I would point out that squeak smalltalk
is free and runs on many platforms, and comes with an extensive object
library.

I would like to see Zocher's SimPCT developed. I learned a lot of PCT by
building models in the original SimPCT, although for my own purposes I now
use forth with object oriented extensions. If a clear specificaton of
SimPCT were written, each of us could implement SimPCT in our favorite
system, and still exchange the SimPCT code.

As I understand it, Delphi is a Windows only product, with no Mac or
Linux/Unix availability. The free Gnu pascal has some Delphi extensions,
but I don't know how compatible it would be

Samuel

···

--
Samuel Spence Saunders, Ph.D. | If man chooses oblivion, he can go right
saunders@gwtc.net | on leaving his fate to political leaders.
ssaunders@olc.edu | If he chooses Utopia, he must initiate an
                                > enormous educational campaign-immediately
                                > R. Buckminster Fuller

13

[From Bill Powers (2002.07.01.17339 MDT)]

Samuel Saunders (2002.07.01.1645 MDT) --

>I would like to see Zocher's SimPCT developed. I learned a lot of PCT by

building models in the original SimPCT, although for my own purposes I now
use forth with object oriented extensions. If a clear specificaton of
SimPCT were written, each of us could implement SimPCT in our favorite
system, and still exchange the SimPCT code.

That is a great suggestion. At the meeting, let's at least write the
specification. Actually, I found the version that just used text files (no
graphics) quite easy and fast to use, but that may be a distinctly minority
view. Actually we could also agree on procedures and functions for carrying
out the various computations in a standard way -- there isn't THAT much
difference between languages, even if some people call them methods.

As I understand it, Delphi is a Windows only product, with no Mac or
Linux/Unix availability. The free Gnu pascal has some Delphi extensions,
but I don't know how compatible it would be.

I guess it is Windows only, though Borland has some other packages, maybe
for Unix or Linux. I saw nothing for a Mac. With your idea, we would just
specify how the various computing blocks would work, and a few other
concepts like grouping, and then let our various programmers make the
system run that way on their own machines in whatever underlying language
turns them on.

Forth with object orientation!!! Wow. I think you're either genetically
disposed to write programs in Forth or you're not (something similar
probably holds for OOP). I've heard it called a write-only language .
Remembering the Forth words you wrote last week is impossible -- except for
those specially endowed individuals who find it easy and can't understand
the fuss. I wrote a rather long typesetting program (for stocks) in Forth,
and it just got totally out of hand before it was complete. I ended up in
Turbo C 2.0 .

Best,

Bill P.

14

Unfortunately I'm just beginning to use Java myself. Having been raised on
procedural/structured programming I, too, find OO a bit difficult to think
in. In fact, the first program I ever wrote was on punch cards. Having
said that, I'm starting to use Java professionally almost daily. And I have
some experienced resources here I can fall back on. At the very least I can
get questions answered.

Steve O

···

-----Original Message-----
From: Richard Marken [mailto:marken@MINDREADINGS.COM]
Sent: Monday, July 01, 2002 2:34 PM
To: CSGNET@listserv.uiuc.edu
Subject: Re: Conference, Modeling

[From Rick Marken (2002.07.01.1430)]

Stephen O'Shaughnessy wrote:

For my 2 cents I think a computer language used for modeling should be
object oriented. I vote for Java. It's inexpensive (Free unless you want

a

nice IDE) and runs on many platforms. I've never learned Pascal, is

Delphi

OO? I've never learned Smalltalk either, but I would be willing.

I second your vote for Java! It's not the OO aspect that matters to me. It's
the "_free_ that matters to me. As does the fact that it's cross platform,
easily accessible to everyone and easily distributed on the web.

There are things about Java that are difficult (and annoying) for an old
fashioned procedural programmer, like me). But these can be easily worked
around. If the goal is to allow a network of highly dispersed people to
develop
and share models of behavioral phenomena, then Java is the way to go.

I think what we need is someone who can teach the basics of Java as clearly
and
simply as possible, with an emphasis on using Java for modeling and,
perhaps,
interactive experiments (like mine on the net). I'm not the one who can do
that, though. I struggle along with Java as best as I can. But maybe you
are?

Best

Rick

--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

15

[From Rick Marken (2002.07.02.1040)]

Bill Powers (2002.07.02.1036 MDT)--

I've programmed in Java, too, but like a blind man in the dark. More to
learn than I wanted to take the time for, and no experts in the background.
Nobody writes manuals any more -- a generation of illiterate programmers.
My impression was mainly of its SSSSLLLLOOOOWWWNNNNEEEESSSSSSSS. I don't
know how that will work out in simulating an analog computer in real time,
where you want to iterate all the blocks at least 100 times per second and
preferably 1000. For running solutions fast while adjusting parameters, you
probably want to go 100 times real time, too. Can that be done in Java?

Java is slow because it has that whole Java machine to go through. But I've
found that it's fast enough to do some reasonable modeling, animations (such as
the ball catching model) and interactive experiments.

It seems to me that Java is more than fast and powerful enough (especially on
the newer, affordable Gigabyte speed processors available) to support the kind
of modeling that would be involved in _teaching_ modeling. I don't believe a
beginning modeler is going to be very concerned about the fidelity of the
digital representation of the analog system. I think it's an important issue and
it should be part of learning how to build models. But that just a little part
of modeling and there are many other important things to learn about computer
implementation of models that Java supports just fine. And Java makes it easy
for a distributed group of "students" to "show their work".

So while Java may not support the needs of an advanced modeler, like yourself, I
think it supports the needs of beginner and intermediate modelers (like me)
just fine.

Best regards

Rick

···

--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

16

[From Bill Powers (2002.07.02.1036 MDT)]

Steve O'Shaughnessy (2002.07.02) --

Unfortunately I'm just beginning to use Java myself. Having been raised on
procedural/structured programming I, too, find OO a bit difficult to think
in. In fact, the first program I ever wrote was on punch cards. Having
said that, I'm starting to use Java professionally almost daily. And I have
some experienced resources here I can fall back on. At the very least I can
get questions answered.

I started that way, too, with an IBM 650. Gawd, that was 40 years ago.
SOAP: Symbolic Operation (?) Assembly Program. Programs that were 200
cards long. Data that occupied 2000 cards, a whole box. Two thousand word
10-decimal-digit memory on a rotating drum. Punchity-punchity-punch.

I've programmed in Java, too, but like a blind man in the dark. More to
learn than I wanted to take the time for, and no experts in the background.
Nobody writes manuals any more -- a generation of illiterate programmers.
My impression was mainly of its SSSSLLLLOOOOWWWNNNNEEEESSSSSSSS. I don't
know how that will work out in simulating an analog computer in real time,
where you want to iterate all the blocks at least 100 times per second and
preferably 1000. For running solutions fast while adjusting parameters, you
probably want to go 100 times real time, too. Can that be done in Java?

Blind man in the dark? Come to think of it, what would the blind man care?

Somebody tell me if my "signature" file is working.

17

[From Bruce Nevin (2002.07.02 14:42EDT)]

Bill Powers (2002.07.02.1036 MDT)–

Nobody writes manuals any more – a generation
of illiterate programmers.

Huh?
O’Reilly does a good job on about everything they touch:
http://java.oreilly.com/
The originators at Sun have lots of resources:

For example:

Here’s what Amazon says are their most popular titles for java
programming:

  1. Beginning Java 2 SDK 1.4 Edition – by Ivor Horton; Paperback
    Our Price: $34.99

  2. Effective Java Programming Language Guide – by Joshua Bloch;
    Paperback
    Our Price: $39.99 – Or buy used from $17.28
    This is leaving aside Java for Dummies and The Complete Idiot’s
    Guide to Java 1.2     (titles that are available).

     /Bruce
···

At 10:56 AM 7/2/2002 -0600, Bill Powers wrote:

18

[From Bill Powers (2002.07.02.1718 MDT)]

Bruce Nevin (2002.07.02 14:42EDT)--

Nobody writes manuals any more -- a generation of illiterate programmers.

Huh?

O'Reilly does a good job on about everything they touch:
http://java.oreilly.com/

What I mean is that programmers seem to rely on code examples rather than
using the facilities of ordinaty language to construct orderly
explanations. But the problem I have is probably my own -- I couldn't get
anywhere with Sun's Java Development Package, and the sample chapter from
O"Reilly's book for Java beginners didn't tell me anything I wanted to
know. Maybe the first chapter is better but I can't work up any enthusiasm
for spending $45 to find out.

I don't want to rant about programming languages. Samuel Saunders' idea
bypasses all that.

Dag has reminded me that an early version of Simcon may be on the CD-ROM he
distributed. I'm going to check.

Here is what my signature file was supposed to look like:

Best,

Bill P.

"Pay no attention to that man behind the curtain"
The Wizard of Oz

Best,

Bill P.

"Pay no attention to that man behind the curtain."
The Wizard Of Oz

···

At 10:56 AM 7/2/2002 -0600, Bill Powers wrote: