[brlug-general] OT: Torvalds on VB

[Chopin Cusachs]

As an old card thumper with more error messages to his
debit than he can ever count, I have some impressions.

What is good or bad in programming is dependent on
what you need to do.  I have solved some real-time data
acquisition problems where a multi-tasking operating
system's services were simply to slow to respond, so we
had to lock some files in memory and write our own, faster,
read/write routines in assembler.

For other, number crunching problems, we had to be
concerned with the efficiency of loops, to minimize the
number of read/write operations, even to the extent of
constructing and managing our buffers.

In the old days of the IBM 650, in the late 1950s, we
would calculate where the read/write heads would be on
the drum after an operation to select the location where
the result would be stored.   Going from explicit op codes
and addresses to a language with symbolic addresses
and mnemonic operations speeded up getting results or
crashes, but deprived us of altering a bad instruction from
the keys.  A real assembly language helped us to find
and fix errors more efficiently.  It was good training.

But I have always believed that structured thinking is far
more important for a programmer than structured
languages.   I retired before object oriented programming
became popular, so I need to refrain from commenting on
it. But I note that VB seems limited to Windows and to
applications where efficient code is much less important
that getting something that works.  I, too, have been
grateful for graphic routines I could call from FORTRAN.
I even wrote a few .

Back when I was in the oil patch, we were concerned
with extracting the maximum useful information from an
instrument that produced 250,000 measurements per
sample.  We used principal component analysis to
search for significant combinations and to get rid of
false correlations from rounding error.  Got a big program
almost finished when one of the old timers much higher
up wanted to know why we didn't use the general stat
package the company had purchased.  I said it wouldn't
work, but the fellow insisted on a test.  We were about
a third way through reading the first data set when the
program barfed and said too much data, cut it back.  The
man said "Why not?"   I said the purpose of the project
was to see if some non-intuitive linear combinations of
variables might be significant.   He shook his head.
Don't think he ever understood.   His eyes glazed over
when I tried to explain the math to him.

But I did turn out some interesting titles, like "Minimum
Multiplication Matrix Manipulations in Marginal Main
Memory" and "A Generalized Gnomonic Projection of
the Spheroid on the Sphere."   The latter saved my
employer from tying up a big mainframe complex for
a week or so.   Ah, what wit and wisdom lies buried in
company archives.  We produced results in FORTRAN,
and the code was almost completely independent of
operating system.

Choppy


At 01:22 PM 10/10/06, Dustin  wrote:

>I tend to agree. As far as VB and its impact (good and bad), I think
>it's really hard to understate it.