| Sunday,
April 1, 2001 In general I do not do
April Fool jokes.
 We
are pleased to hear from Talin after too long an absence from Chaos Manor
mail.
On Python:
Jerry,
My congratulations on choosing a very wise course of
study. Python is, in many respects "Visual Basic done right".
It's a well-organized, well thought-out language that is both powerful and
easy to learn. And, unlike Perl which is it's nearest competitor, the
resulting code is highly readable.
Python has one feature which is hard to get used to,
at least for experienced programmers. In most languages such as C++ and
Java, the block structure of the language is specified using curly braces
{}. However in Python (as in some other scripting languages such as
Hypertalk), the block structure is specified using the level of
indentation. That's right, white space at the beginning of the line is
syntactically significant. This drives long-time coders nuts, at least for
about a week. Then you get used to it, and you don't notice it any more.
(Especially if you use an editor that knows about Python syntax, like IDLE
or the Emacs Python mode.) I resisted learning the language a long time
because of this, and I'm sorry I waited so long. It's a really nice
language overall.
Python is getting a lot of interest in the games
industry (BTW, I should mention I'm back to doing online games, I was just
bored working on e-commerce web application servers).
Traditionally, game developers have created
specialized game scripting languages which are embedded within the game
engine, so that new scenarios and game rules can be added without going
through a lengthy and cumbersome build process. This has benefits for
servers, clients, or standalone games. My own scripting language, SAGA
(Scripts for Animated Graphic Adventures) is an example of this. However,
it would be nice to be able to leverage an existing language, especially
if it could be easily embedded within the C++ code, and it could be
extended to provide "game-like" functionality. And using an
existing language allows you to leveage a large body of existing libraries
and application code. Want to parse XML, or add HTTP downloading of new
artwork and game updates? No problem, just use an existing library and add
it to your product.
Both Python and Java are obvious candidates for
this, because they are both very easy languages to "embed"
within a larger C++ application, and they both have large archives of free
code on the net. At first Java seems the obvious choice (and it's a choice
that a lot of game developers have made) because Java code runs
substantially faster than Python, due to the advanced "Just in
Time" (JIT) compilation technogy of many Java virtual machines.
However, Java has a serious performance problem calling C++ code -
whenever you transition between interpreted code and native code, the Java
VM has to "lock down" the garbage collector, and in some cases
it will actually have to move arrays around in memory in order to make
them accessible to the C++ code. These operations take a lot of time,
resulting in a "speed bump" whenever you call a native function.
Thus, Java/C++ hybrids tend to be designed so that there is very little
interconnection between the Java and C++ code, so that each part can run
relatively independently. This kind of design approach greatly reduces the
benefits of having a scripting language - the ideal architecture is one in
which just about every object in the application is potentially
scriptable. So Python, even though it is potentially slower, actually ends
up being faster in some cases, and allows a much more
"integrated" and fine-grained scripting solution.
Also, because Python is weakly-typed and more
"late-binding" than Java, it is typically more suitable for
rapid prototyping. I have observed that the more dynamic a language is -
that is, having dynamic types, late-binding semantics, and so forth - the
faster you can write programs in it. I can write Java code twice as fast a
C++ code, and I can write Python code twice as fast as Java.
And in truth, the performance difference between the
two languages is really not so important given that C++ code is doing the
heavy lifting anyway. For example, I wrote a little 3D polygon editor demo
using Python. The actual 3D drawing was all being done by OpenGL, using
the PyOpenGL library which provides Python equivalents of all of the
OpenGL calls. 95% of the execution time is spent in the OpenGL code, so
the performance of the scripting language is really unimportant.
Another nice Python library is wxPython, which is a
Python interface to the popular C++ wxWindows library. wxWindows is a
cross-platform GUI library, similar in concept to Tk or Java's Swing.
However, wxWindows has a number of advantages over these other libraries:
1) The widgets, such as buttons and scrollbars, are real
"native" widgets, with thin cross-platform wrappers around them.
I'm continually frustrated whenever I run a Java application under
Windows, because the mouse wheel doesn't scroll the window. The
"synthetic" widgets used by Java are immune to all of the
various platform-specific enhancements related to alternative input
devices and accessibility for disabled users. The widgets used by
wxWindows don't have this problem, since they are native widgets. 2) Many
cross-platform libraries take a "least common denominator"
approach - they provide only the features that are the logical
intersection of the features of the individual platforms. wxWindows takes
the approach of bringing all of the platforms "up to par" with
each other, so that you get the full feature set regardless of whether you
are running on Mac, Linux, or Windows.
Using Python and wxPython together allows you to
quickly make simple GUI-based applications that run on a large number of
platforms. Add in OpenGL, and you have the ability to construct
sophisticated 3D editing, modeling and visualization applications
relatively quickly compared to C++ development. It takes some time, of
course, to learn the various APIs, but all of the components I have
described are extremely well documented, with extensive tutorials, samples
and reference materials on how to construct an application. And all of
them have a strong community of enthusiastic supporters, which is an
important factor in choosing a development environment.
-- Talin
"I am life's flame, respect my name,
my fire is red, my heart is gold.
Thy dreams can be, believe in me,
if you will let my wings unfold!"
-- Heather Alexander
http://www.explorati.com
http://www.sylvantech.com/~talin
Python is
an interpreted language so it is not for you if you want to compile. It is
certainly fast enough for anything I want to do except long simulations.
My impression is that Python is to Perl as Pascal is
to C. I can understand my Python scripts a week from now, while I have a
lot of C code that takes as long to understand as it took me to write it a
year ago.
Nesting is determined by indentation, which some
find annoying but which I find to be elegant and simple. Just don't mix
tabs and spaces.
Henry Harpending
Henry Harpending harpend@xmission.com
Dept. Anthropology, University of Utah
With modern computers interpreted vs. compiled is
no longer such a large issue; and of course any popular language can
generate a compiler. Pascal for a long time ran on a P-Code as did many
commercial Basics in the early days of small computers. Eventually Pascal
compilers were written in Pascal (the original Pascal was a teaching
language and didn't even have an interpreter much less a compiler). I make
no doubt that if Python catches on someone will write a Python compiler...
And of course the very picky Pascal compilers
with range checking were one of the reasons for Pascal's success before
the lack of I/O libraries and file structures overwhelmed it. Wirth went
on to Modula 2 and Oberon, and Pascal was neglected except by Borland.
Alas.
To me the important thing is to be able to see
what I have done 5 years later. This is why I will never be able to
program in C. Of course as I get older this may be a moot point...
I will probably collect all Python related
materials onto a report page, but let's let things go ahead in mail for a
while. Thanks.
On the Power crisis:
Dr. Pournelle,
I wondered if you had seen these two articles
regarding the power crisis in CA. Upon reading the first I felt minor
alarm, but gave incompetence greater weight than malice. However, after
reading the second, I have reversed my position.
This fragment taken from the opening paragraph of
the second article frames the series well: ". . . The Third Way and
Loretta Lynch, Commissioner of the California Public Utilities Commission,
which orchestrated the state energy crisis."
http://www.worldnetdaily.com/news/article.asp?ARTICLE_ID=22009
http://www.worldnetdaily.com/news/article.asp?ARTICLE_ID=22245
This is scary stuff! I certainly hope it is not
true, but given the debacle of the last 8 years of Clinton leadership I'm
afraid it may be so. . .
A side note: The statistics put forth in US
electricity cost map in the MSNBC article posted early ( http://www.msnbc.com/news/512303.asp
)are a typical example of the press either misinterpreting or
misrepresenting facts. The numbers given in that map regarding avg US
electricity costs are meaningless. They don't take into account the fact
that all US households don't use the same energy source for heating and
cooling or that the heating and cooling loads are quite different across
the nation. The meaningful comparison would have been kwh cost, not avg
annual bill. The comparison would have been dramatically different had
that been the case.
Brian Innovate... relentlessly.
I don't know the World
Net Daily. From the layout it looks a bit like The Washington Times Insight
magazine, which is one of those odd operations ultimately owned by the
Reverend Moon. I know some of the editorial people at Insight (at one time
edited by Arnaud de Borchgrave) and they are not Unification members nor
are they controlled or controllable by Moon, so this isn't a problem with
me, but it is the kind of thing people ought to know. I subscribe to the
paper version of Insight and have for years.
IN the case of World Net Daily I don't know a
thing: the masthead doesn't reveal anyone I know, which given my memory
isn't surprising, but given where WND is published it's hardly likely to
be part of Moon's group. Given the advertisements it doesn't look to have
a large liberal readership...
Thanks for the pointer to it. Looks like an
interesting news source. Now regarding the power crisis:
I find the whole subject depressing, but then I
did when I had all those debates on how conservation wasn't going to do
the job and we needed new energy sources. And note that if the Washington
Public Power Systems reactors had been built at what was then said to be
inflated costs for needless systems, the low snowfall (which is what they
were designed to compensate for) would not be a problem now.
Conservation does not expand an economy. It may
be necessary or even desirable, but it is usually best achieved by
allowing products and services to reflect their true market costs, thus
encouraging resource substitution and conservation where those are
appropriate.
Incidentally why does anyone expect a bureaucracy
to act in any but bureaucratic ways? Power corrupts.
I wrote about the fallacies of the Club of Rome
and the Forrester/Meadows models of doom in my A STEP FARTHER OUT more
than 20 years ago. No one listened then either.
Subject: Dual CPU boards
Perhaps you'd be in a position to know this...I'm in
the position where I have to configure several workstations in the near
future....AMD has been promising its chipset for dual Athlons
forever...Intel must be working on a DUal PIV soulution....do you know if
either of these are likely to show sometime soon? Any help or direction to
info would be greatly appreciated
Thank you Louis Markoya
There was a dual Athlon board in the AMD booth at
WINHEC. NO ONE was allowed to look inside the box (and I mean no one
including Carl Stork; I was there when he asked, as I did) and no
announcement schedule much less shipping date was announced.
Jerry
Many thanks I have come to find out that AMD will
"offically" announce their Tyan dual Athlon motherboard on June
4th (right from Tyan) with production quantities available 1 month
after...and it's supposed to be an absolute screamer (and very expensive)
Thanks again Louis
Interesting. They didn't tell us that at WINHEC.
The April issue of The Atlantic Monthly contains an
article by Steve Olson ["The Genetic Archaeology of Race"] which
in essence denies the concept of race. Olson writes that "people are
too closely related -- and have mixed too much throughout history -- to
differ in fundamental ways," adding that we "have to remember
just how small the genetic differences among groups are."
But simply because human populations share roughly
99.9 percent of their genes does not mean that there are no significant
differences. Since there are approximately 30,000 to 40,000 human genes,
each made up of hundreds of thousands of base pairs, a 0.1 percent
difference is actually quite a large degree of variation, even though the
layman may not realize this. The number of genes involved in racial
differences may be very small, but that does not mean that the effects are
necessarily small.
Even if all of the genes other than one were
identical between two groups, one difference within that one gene would be
sufficient to produce large group differences, since it is not the
quantity of genetic difference that is important but the nature of the
difference.
For example, current evidence suggests that all of
the sex differences between men and women are the result of just one gene.
This would mean that men and women are almost 100 percent genetically
identical, yet no one suggests that sex is a mere social construct. It is
worth noting that a butterfly and the caterpillar from which it developed
are 100 percent genetically identical. The genes do not change; the
enormous differences between caterpillar and butterfly result from the
activation of different genes at different times. Olson seems not to
realize that the fact of shared genes is not nearly as important as is the
timing involved in adding their products to the mix. Even a small number
of genes, each blinking on and off over time leads to an incalculable
number of possibilities.
Breeds of dogs are analogous to human races, and it
is likely that they are as close genetically as different races of humans,
but there is no doubt that these subtle variations result in significant
differences in appearance, behavior, and intelligence. Ask scientists
working on the Dog Genome project about how few genes separate Golden
Retrievers from Dachshunds. Cattle breeders know that there only a few
genetic differences between Guernseys and Longhorns, yet these breeds are
immensely different, and the difference is not cultural...
Name withheld by request
This is one of those topics it is impossible to
write about: scientific discussion of such matters is taboo. Of course the
Christian view is that we are all wretches and thus the individual
differences are unimportant; and the Declaration of Independence proclaims
legal equality (but once again invokes a Creator, leaving the mechanisms
of equality as a mystery of faith).
Absent religious reasons to assert equality we
are confronted with evidence. Every farmer knows that Guernseys don't make
good steers while Longhorns aren't useful as milchcows, and every child
knows the difference between a Rotweiler and a Cocker Spaniel is not
cultural. The question is what ought a republic to pay attention to? And
if we decide we will ignore race -- 'equal protection of the laws' only
this time we really mean it -- then we do have any business collecting or
being legally aware of race?
Clearly all races produce both bright and stupid
people, and racial statistics deal in averages, not with individuals. We
can make predictions about collections of people that we simply cannot
make about individuals. Now what?
Thorny questions, but putting on scientific
blinders is not likely to solve them.
And while I am at it:
On TMI
I read your remarks about the incident. Actually,
the TMI accident seems to have been a much bigger one than Cernobyl. What
kept it from becoming a "GAU" was mostly the solid construction
of the building that contained the melting core. Am I right in this, or
could you correct me?
Sincerely, K.B.
Construction had a lot to do with the difference
in the events, but design was the important factor. Chernobyl was a
"positive void" reactor, which ejected a large part of its
nuclear inventory into the atmosphere following operator errors.
Positive void reactors (fail dangerous design;
fail disastrous design) cannot be licensed in the US. Edward Teller
had that written into US public law a long time ago. There are none and
none can be built here.
Containment and the like were important in making
TMI a trivial event for public safety (financially it was another story).
Had you tried to get the largest possible dose of radiation from TMI by traveling
around the perimeter with a counter and placing yourself in the hottest
places you found at all times, you would have managed about 2 chest x-rays
worth of radiation during the entire event. Trivial.
I know that Sternglass and Gofman are still
saying what they always have said, that any radiation is horrible, but in
fact the data are now in about hormesis. And interestingly, cancer rates
in Denver where you get more natural radiation than in Los Angeles, are
lower than for Los Angeles. And all the data are in that direction. Small
amounts of radiation don't harm, and may even do you some good. The
Swedish Army has some similar data (there are several places with natural
high radiation and Sweden has universal conscription so the studies aren't
t0o expensive). So TMI was a trivial event, healthwise.
It was also a good test in that everything that
could go wrong did.
Japan is about 40% nuclear powered now;
interestingly most of their reactors were built by US companies. France is
70% nuclear using French nuclear engineering.
TOP
|
