Personal History of Computers
Copyright 1998, 2001 by Ronald B. Standler
Table of Contents
1. Mainframe computers
2. Handheld calculators
3. Hewlett-Packard desktop computers
4. Apple personal computers
5. IBM PC
Abandoning loyal customers
Hewlett-Packard scientific/engineering workstations
IBM Professional FORTRAN
AST Cupid Architecture
I have used computers since 1968. I am continually struck by the incredibly
fast pace of development of computer technology, particularly hardware,
which causes a change in attitudes of programmers and other users.
While this document does contain some facts about history of
computing since the mid-1960s, this is mostly an essay from a personal
perspective of how the development of computers affected my career.
This essay is a bit disorganized, because discussion of computer hardware
and software are intertwined, and also the theme of abandonment of loyal
customers by manufacturers of software and hardware is intertwined with
the history of the software and hardware.
The above Table of Contents, plus links inside this document,
should help supply organization. First, I discuss the machines in
chronological order of their appearance in the marketplace, then I discuss
some of the general issues.
I mention prices for some of my purchases, so readers can see how much cheaper
modern machines are than the early machines, a point that I make
later in this essay.
I close with some comments about the poor way manufacturers of
hardware and software treat their loyal customers.
1. mainframe computers
I taught myself how to program computers in 1968. I used a Burrows B5500
mainframe computer at the University of Denver. I asked the Chairman of the
Physics Department if I could have a computer account and he kindly granted
my request. At that time, students studying for a Bachelor's degree in
physics or chemistry were not required to take classes in computer
programming. The attitude among the physics faculty was that computer
programming was a skill that one could learn without taking a formal class.
While it was true that one could teach one's self, there are a number of
important issues in programming style and construction of
algorithms that are unlikely to be discovered during self-instruction.
This attitude among older physicists reflects a "brute force" approach to
programming that is an inefficient use of programmer's time
(it is easy to write code that is difficult to debug) and also an inefficient
use of computing resources.
Students studying for a Bachelor's degree in engineering during the 1960s
were required to take a class in computer programming.
If one looks at textbooks in computer programming from
the late 1950s through the mid-1960s, many of them were written by
chemical engineers. Chemical engineers were eager users of computers
to solve difficult problems in process optimization that involved sets
One of the considerations in writing programs for mainframe computers
during the 1960s was efficiency. It was much faster to calculate
A + A
than to calculate
2.0 * A
Similarly, it was faster to calculate
0.5 * B
than to calculate
These considerations are still important for code inside nested DO loops,
code which might be executed more than 105 times during one run
of the program.
In other applications, these considerations became obsolete in the
1980s when most programs were run on desktop (i.e., personal) computers.
When one owned the machine, instead of paying for computation time by the
second on a mainframe, it really did not matter if the computation took
a few extra minutes to perform.
The first computer language that I learned was ALGOL.
This was a popular programming language in Europe, but was uncommon
in the U.S.A. Programmers in the U.S.A. usually used COBOL for
business applications and FORTRAN for scientific or engineering applications.
COBOL and FORTRAN were developed IBM, who dominated the market for
mainframe computers in the U.S.A. IBM's dominant role in the hardware
market lead to a similar domination of the software market,
an event that was repeated in the 1980s when the desktop computer
compatible with the IBM-PC propelled Microsoft to domination of the
ALGOL was a block structured language (i.e., a group of commands could
be set between a BEGIN and END statement and treated structurally
as one command). FORTRAN was not a block structured language:
most long FORTRAN programs included GO TO statements, so that
execution hopped around the code, like a frog.
The official ALGOL language had a weak input/output commands.
The version of ALGOL that I used had been extended by Burrows to include
FORMAT, READ, and WRITE statements similar to what was used in FORTRAN, which
greatly improved the ALGOL language.
In 1970, I realized that most computers in the U.S.A. did not have
an ALGOL compiler, so I began to teach myself FORTRAN IV.
I was immediately impressed by the awkwardness of FORTRAN code.
I began to write pseudo-block-structured FORTRAN by defining a
REAL*4 A, DX, LASTX, X
ENDX = X .GT. A
IF (ENDX) LASTX = X
IF (ENDX) DX = DX * 0.1
IF (ENDX) GO TO 100
IF (.NOT.ENDX) GO TO 200
When FORTRAN77 was introduced, one could write
IF (X .GT. A) THEN
LASTX = X
DX = DX * 0.1
GO TO 100
GO TO 200
With the introduction of FORTRAN77, the formerly clunky FORTRAN language
became convenient for writing decent code. However, computer
science experts were strongly prejudiced against FORTRAN, because
of defects in earlier versions of FORTRAN.
Such discussions remind me of arguments about whether,
for example, the English or German language is superior.
The truth is that one can express oneself well in either English or
German, just as one can write computer programs to do useful work
in FORTRAN. If you think I am unreasonable in clinging to FORTRAN,
consider how you would feel if a professor of languages told
you that, because Esperanto is "better" than English, (1) you must
learn Esperanto and use Esperanto in the future, and (2) all of your
past writing must be translated to Esperanto.
I last used a mainframe computer in the early 1980s. By the mid-1980s,
desktop computers were adequate for most scientific computations,
except for weather forecasting, astrophysical calculations, and other
problems that involved solutions of differential equations for over
large amounts of space and time. A desktop computer was, of course,
slower than a mainframe computer. However, use of a mainframe computer
was expensive: one paid for each second of computation time. On the other
hand, one owned the desktop machine, so aside from the trivial cost
of electricity there was no additional expense when a desktop computer
trudged for hour(s) during a numerical simulation.
2. Handheld calculators
In 1968, Hewlett-Packard produced a desktop calculator that would do
not only arithmetic operations, but also common transcendental functions.
That sounds trivial now,
but it was a big improvement to people with a large amount of
laboratory data to process. Before the introduction of such calculators,
we used slide rules for most calculations. A slide rule could give two,
maybe three, significant digits. Those who needed more accuracy used
tables of functions in books, such as Burington's Handbook of Mathematical
Tables and Formulas that was published by McGraw-Hill or the tables
published by Chemical Rubber Company (CRC).
In January 1972, Hewlett-Packard introduced the first handheld calculator,
their model 35. Thirty months later, more than 75 000 of these
calculators had been sold, mostly to scientists and engineers.
Anyone who thinks scientists and engineers are resistant to change should
consider this fact: less than five years after the introduction
of handheld calculators, slide rules and tables of logarithms were
no longer seen on college campuses.
In 1973, Hewlett-Packard introduced the model 45 handheld calculator.
When the price was decreased to $ 325 in July 1974, I purchased one.
It used Reverse Polish Notation for the sequence of numbers and operators
and had a stack of four storage registers (called x, y, z, t),
plus nine addressable (STO # and RCL #) storage registers.
Numbers could be displayed in either scientific notation
(range from 1.0 × 10-99 to 1.0 × 10+99)
or in fixed notation. This calculator was not programmable:
the user needed to key in all numbers and operators and the calculator would
display the result after each operator was pressed. This calculator could do:
This calculator used three nickel-cadmium batteries and could operate for
three to five hours without being charged.
- ex and 10x
- logarithms in both base e and base 10
- all of the trigonometric functions, in units of either
degrees/minutes/seconds of arc, radians, or grads
- square root and x2
- rectangular to polar conversions
- factorial operator
- averages and standard deviations of N entries
I tested this calculator by comparing its result to the values in the
comprehensive tables in the National Bureau of Standards
Handbook of Mathematical Functions by Abramowitz and Stegun, so I
was satisfied that the calculator was accurate. I can remember a physics
professor in 1974 who asked me if my calculator was accurate.
I told him it had nine significant digits.
He punched in an integer number that he chose at random, took the sine,
and then took the inverse sine. He was amazed to find his original
number on the display! I still remember his giggle of delight,
like a child with a neat toy.
3. Hewlett-Packard desktop computers
In 1972, Hewlett-Packard introduced their model 9820A desktop computer that
was similar to the model 45 calculator, but could be programmed to
execute a series of steps and had a much larger memory.
Their machines during the 1970s all had a light-emitting diode display
that showed only one line of code or one numerical result
and had a thermal printer that printed on a narrow strip of
paper, like the receipt at a grocery store. Their computers could
also control a pen plotter, if one wanted graphical output.
In 1976, Hewlett-Packard introduced the model 9825A desktop computer
with a proprietary 16-bit microprocessor. This was the first model of
Hewlett-Packard desktop computer to have a regular QWERTY keyboard,
instead of function keys like a calculator. The 9825 computer was also the
first to use Hewlett-Packard Language (HPL),
a curious blend of programming concepts from handheld calculators,
assembly language, ALGOL, and FORTRAN. HPL and the operating system was
stored in semiconductor read-only memory (ROM), so no separate step of
compilation was needed. Numbers inside the machine were stored in binary-coded
decimal (BCD) format, so there was no distinction between integer and
real numbers. HPL was the easiest and most convenient computer language
that I ever used. Unfortunately, HPL disappeared
in 1985, partly because the integral use of BCD
in HPL was inefficient use of memory space, and partly because of the
industry's trend toward standard languages like PASCAL, BASIC, and C.
I was so excited about the possibilities of having my own computer that,
in February 1981, I spent $ 8200 and purchased a Hewlett-Packard model 9825T
desktop computer for my home.
The model 9825T came with 64 kilobytes of
semiconductor memory (RAM), the maximum that its microprocessor could
address. I wrote hundreds of programs for this computer
and I never came close to using all of the memory, but then I did not
use large arrays in any of my programs for this machine.
I spent another $ 3000 in Feb 1981 and purchased a Hewlett-Packard model
7225 plotter, and an IEEE-488 interface for the 9825 computer.
I used this plotter to make overhead transparencies that I used in my
lectures on analog electronic circuit design and electromagnetic field theory,
as well as plots on paper for my publications.
I really enjoyed using the Hewlett-Packard model 9825, but I wanted
a display that would show more than one line of code and I wanted more
In Aug 1984, I spent about $15500
and purchased a Hewlett-Packard model 9836 (later called a model 236)
with a 15 Mbyte hard disk for my personal use.
This computer contained a Motorola 68000 microprocessor with
640 Kbytes of RAM. When the model 9836 was first introduced in 1982,
it far outperformed any IBM-PC: the model 68000 microprocessor had
32-bit data and address registers and had an 8 MHz clock rate.
I later upgraded my model 9836 to contain 1664 Kbytes of RAM.
Later, I purchased two Hewlett-Packard model 310 computers
to transfer data from digital oscilloscopes in my laboratory at
The Pennsylvania State University, and to analyze that data.
The Hewlett-Packard models 9836 and 310 used the same operating system
and same BASIC programming language.
4. Apple personal computers
There were a variety of desktop computers sold during the 1970s for use at
home by now long-forgotten companies. To use these machines,
one had to be not only a computer programmer, but also an enthusiast
about electronic hardware, as they were not user-friendly.
A real revolution began in 1977, when the Apple II desktop computer was
introduced. The Apple II had a model 6503 microprocessor with between
4 and 48 kilobytes of semiconductor memory (RAM). The first machines
had a cassette tape drive for program storage; in 1978 a
5¼ inch floppy disk drive was added.
In 1979 VisiCalc spreadsheet software became available for the Apple computer.
In 1984, Apple introduced the Macintosh computer, which used a
Motorola 68000 microprocessor running at 8 MHz
and had a 3½ inch floppy disk drive.
In March 1994, Apple began to use the PowerPC microprocessor from Motorola.
The latest Apple machines (with the PowerPC microprocessor) and current
Apple operating system will still run old software that was designed
for Macintosh computers that used a 680X0 microprocessor.
I commend Apple for this backward compatibility, which is rare in the
computer hardware/software area.
It is also worthwhile mentioning that "plug and play" was a reality for
the Apple Macintosh more than ten years before that concept came to
PC's using Windows95. Further, there is no Y2K problem on Apple machines.
A detailed history of all Apple computers is available on the web at
During the 1980s, I had never seriously considered purchasing
an Apple computer, because Apple's advertisements had slogans such as
"a computer for the rest of us", which implied that their computers
were toys aimed at children and adults who were computer-illiterate.
I was proud of my skills in computer hardware and
software and I did not want to own such a toy.
However, in May 1992, I purchased a computer-controlled synthesizer
so I could hear some of my arrangements and compositions.
At that time, the best music notation software was primarily
developed for the Macintosh platform, so I purchased an Apple IIci computer
to control my synthesizer. There are two reasons why professional
music software was principally a Macintosh product:
- The Macintosh operating system was better suited to graphical notation
(i.e., staves of music) than early versions of Windows.
- Most musicians were not knowledgeable about
computer hardware and software. The fact that the Macintosh operating
system was much easier to operate than the Windows operating system made
the Macintosh a more comfortable environment for musicians; the
Macintosh was much less intimidating to musicians than Windows.
I immediately found the Macintosh to be superior to the PC in many ways.
For example, there are standards to make all applications on the
Macintosh use consistent commands and keystrokes, which makes software easier
to learn to use. As mentioned above, "plug and play" worked reliably on
the Macintosh 10 years before Windows.
In December 1994, Apple licensed its computer technology to manufacturers
of clones, in an attempt to expand the market for the Macintosh operating system.
One of the clone manufacturers, Power Computing,
made computers that offered more features, were faster, and were less
expensive than Apple's own computers. Power Computing sold build-to-order
computers directly to users, while Apple traditionally sold computers only
through dealers. Unable to stand the competition,
Apple announced on 2 Sep 1997 that it will purchase Power Computing
and Apple decided not to renew its licenses with other manufacturers of Mac clones.
I have been very happy with a Power Computing machine that I purchased in
May 1996, now I own an orphan. <sigh>
Apple's hardware has a reputation of having higher performance than
IBM-compatible PC's, but Apple's hardware was also more expensive than
IBM-compatible PC's. In the late 1990s, Apple made
several changes to their hardware:
These changes allowed Apple to take advantage of peripherals that
were produced in much larger quantities for the IBM-compatible
personal computer market, thereby being less expensive and giving
Apple's users a wider choice of peripherals.
- Around 1995, Apple abandoned their internal proprietary bus (NuBus)
and began using the PCI bus that was used in IBM-compatible personal computers.
This change meant that Apple computer could use the less expensive
graphics cards that were manufactured in large quantities for IBM-PC clones.
- Around 1997, Apple began using hard drives with the IDE/EIDE/ATA
interface that was commonly used in IBM-compatible PCs since
the late 1980s. Previously, Apple used hard drives with the SCSI
interface, which were somewhat more reliable than the IDE interface,
but also more expensive than the IDE interface. The SCSI interface could
accommodate seven different devices on one interface, while the IDE
interface can have only two devices per interface.
- Around 1999, Apple abandoned its proprietary Apple Desktop Bus (ADB)
that was used to connect a keyboard and a mouse or trackball to
the Apple Computer. Instead, Apple began to use keyboards and
mice that used the Universal Serial Bus (USB).
My favorite software
for the Apple is listed in a separate document.
5. IBM PC
In 1981, IBM introduced its personal computer. Many people believe that the
IBM PC was the first personal computer, because they are unaware
that Hewlett-Packard sold desktop computers more than seven years before IBM,
Apple sold personal computers four years before IBM,
and that there were other earlier manufacturers of personal computers.
The IBM-PC machines used an Intel 8086 microprocessor
that operated at 4.77 MHz.
The IBM-PC/AT machine was introduced in 1984, which used an Intel 80286
In 1987, IBM introduced the PS/2 computer that used an Intel 80386
microprocessor and 3½ inch floppy drives.
Hewlett-Packard scientific desktop computers and Apple Macintosh had
earlier used the 3½ inch floppy disk, but IBM's choice of
this disk format established a new industry standard that persists today.
software for IBM-PC
In 1982, the Lotus 1-2-3 spreadsheet program became available for the IBM-PC.
In 1983, the WordPerfect word processor program became available for the IBM-PC.
The availability of software for the Apple II and IBM-PC began a new tradition in computing:
no longer did a computer user need to be a programmer.
I return to this thought later in this essay.
The IBM-PC used a disk operating system (DOS) developed by Microsoft.
The immense popularity of the IBM PC propelled Microsoft to its dominant
position in the international software industry.
When the original operating system for the IBM-PC, DOS, was developed,
the maximum amount of semiconductor memory was set at
0.64 Megabyte. Now, we would say only 0.64 Megabyte,
but when DOS was developed, even the visionary Bill Gates is rumored to
have said in 1981 that "640 Kbytes ought to be enough for anyone".
In May 1990, Microsoft released Windows 3.0, an operating system
similar to what was used earlier on the Apple Macintosh computers.
The final version of 16-bit Windows, Windows for Workgroups 3.11,
was released in November 1993. Thereafter, Microsoft
abandoned its large
customer base who continued to use 16-bit versions of Windows.
Operating systems have become bloated and slower, although
the hardware is much faster.
My computer in 1991 that used a '386 microprocessor running at 33 MHz
would be ready to run DOS version 5.0 just 25 seconds after switching
on the computer.
My computer in 1998 that used a Pentium II microprocessor
running at 400 MHz would be ready to run Windows NT version 4
about 95 seconds after switching on the computer.
My favorite software
for the PC is listed in a separate document.
Apple vs. IBM-PC
Apple Macintosh computers became common in the desktop publishing area,
because of their excellent graphical operating system and the availability
of software like Adobe's PostScript fonts and Photoshop software.
Apple computers also became common in schools and universities,
because of their ease of use. However, in all other environments,
IBM-compatible PCs became the dominant desktop computer.
In the mid-1990s, Apple had only about a 5% share of the desktop computer
market, nearly all of the remainder was PC's running Windows.
One of the things that caused the IBM-PC to become the most popular
personal computer and the de facto standard was the decision of
IBM to allow other computer manufacturers to produce
compatible machines. By the late 1980s, companies like Compaq
were selling more IBM-compatible machines per year than IBM.
- Apple used a proprietary operating
system that Apple refused to license to clone computer manufacturers
(except briefly during 1995-97). If Apple had licensed
clone manufacturers in the early 1980s, perhaps it could have become
the dominant operating system, instead of DOS/Windows.
- Hewlett-Packard desktop computers sold to the
scientific and engineering market in the 1970s and 1980s also used a
proprietary operating system.
When Apple's advertised their computers in the late 1970s and
early 1980s as a computer for people's homes and elementary schools,
as opposed to offices and laboratories, Apple did itself a big disservice.
Many of my colleagues did the same as I and simply accepted Apple's
characterization of their computers as an easy-to-use machine for
recreation. If businesses and offices had purchased Apples instead of
PCs, there would have been a large savings in the cost of training
people to use the new computers, since the Apple is much easier to use
than a PC.
The availability of software for the Apple II and IBM-PC,
like Lotus 1-2-3 and WordPerfect, began a new tradition in computing:
no longer did a computer user need to be a programmer.
A user could now purchase software and use it without understanding anything
about computer programming languages or hardware. Of course, fluent
use of some software requires the ability to write macros for the application
and for the user to interact intelligently with the operating system.
Starting around 1980, there was an amazing development of computer technology.
Not only did performance increase greatly, but also prices were drastically
decreased. Some of the price decrease was caused by economies of scale,
as more customers purchased equipment, but most of the price
decrease came from advances in manufacturing technology.
- In 1986 one could purchase a 10 Mbyte hard drive for $ 1000.
Such a small disk drive was no longer available in Sep 1998,
but one could then purchase a 3220 Mbyte hard drive for $ 140.
The capacity changed by a factor of 322 and the price changed by a
factor of 0.14, which was a factor of 2300 improvement in the price
per byte of storage. Further, the newer disk drive had much larger on-board
cache memory, smaller disk diameter, and faster disk rotation rate,
so read/writes were much faster on the newer disk drive.
- In November 1981, I paid $ 1000 for a Racal-Vadic external
modem that could send/receive at either 300 or 1200 bits/second.
In January 1998, one could purchase a Hayes external
modem for $ 80 that could operate at data rates up to 33600 bits/second.
The speed increased by a factor of 28 and the price changed by a factor of 0.08.
The electrical engineering profession should be proud of this factor of 350
improvement in performance/price ratio in just 16 years! And this
computation ignores the fact that money in 1997 was worth less than in 1981,
owing to considerable inflation in the intervening years
if inflation is considered the improvement is even greater!
- In January 1982, I purchased a Digital Equipment
Corporation (DEC) video terminal and keyboard, model VT102, for $ 1925.
This was only a dumb terminal, with a monochrome green CRT, to allow me
to use the mainframe computer at the University via
modem from my home. Sixteen years later, one could purchase both
a standalone computer and a color monitor for the same amount of money,
again ignoring the effects of inflation in the intervening years.
In 1965, Gordon Moore, co-founder of Intel, observed that the maximum number
of transistors on one semiconductor chip doubled every 18 to 24 months,
and the cost/performance ratio decreased at a rate of 35%/year.
The first part of this observation became known as Moore's Law, and
it continues to describe progress in the semiconductor industry.
The number of transistors on one chip is proportional to the amount of
memory in bytes in a RAM chip.
Aside from the invention of vaccines and antibiotics in the
medical profession, I would challenge any other profession to demonstrate
a greater improvement in performance/price ratio in such a short time.
This rapid rate of progress in computer hardware becomes more dramatic
when compared with the relative lack of progress in other areas.
For example, apply the second example above to automobiles.
In 1981, an automobile could travel at about 40 meters/second
(about 90 miles/hour in the USA, albeit illegally),
so 28 times as fast would be 1120 m/s, more than three times the
speed of sound. In 1981, an automobile cost about US$ 12,000, so a price
change of a factor of 0.08 would give a current cost of about $ 960.
Therefore, if the changes in performance in the modem market from 1981 to
1998 were applied to the automobile market, in 1998 automobiles would travel
at more than three times the speed of sound and cost $ 960 each.
In 1982-83, I sought a wordprocessor that could do equations conveniently
with a full set of mathematical symbols and Greek letters. I found a little
company in Waltham, Massachusetts, ComputerMart, that had developed a
wordprocessor called Techwriter. Their wordprocessing software did
everything that I needed, so I purchased it in August 1983,
along with their preferred hardware: a North Star Advantage computer
and Xerox-Diablo model 630ECS daisy wheel printer at a cost of $ 7500.
I did not care about the details of the computer hardware or operating
system, as I only used this machine for wordprocessing.
The Advantage had two floppy disk drives: one for the wordprocessing
software and one for the user's documents. There was no hard disk on my
Advantage, although versions with one floppy drive and one hard disk (either
5 or 15 Mbyte) were available.
By 1985, it was difficult to find the 5¼ inch floppy disks
that were required for use in the Advantage (e.g., Dysan 107/2D).
In 1986, I purchased Techwriter software for the IBM-PC,
at a cost of $ 795 for a non-copy-protected version with a spelling
checker. In August 1986, I installed a card in my IBM-PC that would
enable the floppy drive to read the diskettes from the North Star Advantage,
so I could convert all of my documents from the Advantage to the IBM-PC.
In November 1989, I installed version 3.1 of Techwriter,
but I found that it did not work, so I returned to version 3.0 the
next day. The next time I tried to contact ComputerMart, the developer of
Techwriter, I found that they had ceased business.
I was stranded with an orphan wordprocessing program that had a
software driver only for a Hercules monochrome graphics card.
(ComputerMart had written a driver for a color graphics card, but I did not
purchase that driver before they ceased business.)
I had written two technical books and dozens of technical papers in
Techwriter, which I could not easily convert to any standard wordprocessor,
such as WordPerfect, although I could export a plain ASCII text file without
any of the formatting commands and special characters.
Techwriter was limited to the fonts and sizes available in ROM
in the printer, for example the B and J font cartridges for the
Hewlett-Packard LaserJet II printer.
In October 1989, I purchased WordPerfect 5.0 for DOS to get the capability
to switch fonts and font sizes in my documents, which I needed for producing
visual aids for my lectures.
I upgraded to WordPerfect 5.1 for DOS in March 1991 and
promptly began using it as my principal wordprocessor.
I installed WordPerfect version 5.1+ in November 1994, which
had scaleable fonts, a big improvement from the bit-mapped fonts in version
5.1 and earlier.
I have tried WordPerfect versions 6.0 and 6.1 for DOS, but I did not like them,
so I continue to use 5.1+ as my principal word processor.
Nearly all of my word processing operations involve the
following small set of commands:
Some of my documents also use:
- cut and paste a block of text
- copy a block of text
- search for certain words or symbols
- search for certain words or symbols and automatically replace them
with other words or symbols
- import another file into the current document: this other file could
be another wordprocessing document or a file with ASCII text
- spell checking
- header on every page with document title, date of printing, and page number
- attributes: boldface, italics, underlining, subscripts, superscripts
- indent a block of text
- center text on a line
- set tabs
- use Greek letters, mathematical symbols, characters from foreign
alphabets (e.g., é ö ß)
- select font and size
- change line spacing
- export the word processing document as either an ASCII or HTML file
Word processors in the mid-1980s could do most of these operations.
Users might be better served by a simple word processor that would do these
fundamental operations, plus the ability to add
modules to do additional operations that were important to the particular user.
New modules for old software could be released, in the same way that one can
purchase parts for a 10 or 20 year old automobile.
For example, it would be nice in 1997 to add the ability to export a
document as an HTML file to WordPerfect version 5.1+ for DOS, which was
released in March 1994, instead of needing purchasing
the latest version of WordPerfect, which has a different "look and feel".
- table of contents
- references to the page number of a target somewhere in the same document
- equations with long fractions that use more than one line of printed text
- add a column of numbers
- table of authorities (only in a brief for an appellate court)
As an example of how not to do it, highlighting a document in the WordPerfect 5.1
file menu and pressing 6, displays the document quickly without the ability to edit it.
The same command in WordPerfect 6.1 for DOS deletes the document!
As another example of how not to do it, WordPerfect 5.1 for DOS
consumes about 4.6 Mbytes of disk space, not including dictionaries
and macros. The same files for WordPerfect 8 for Windows 95/98
occupy more than 68 Mbytes.
In July 2001, I installed WordPerfect 10 on my Windows 98 computer.
Although I did a custom install of only the wordprocessor and associated
files (e.g., spellchecker), WordPerfect 10 wrote about
205 megabytes of files to my hard drive.
Worse, my WordPerfect 10 upgrade disabled WordPerfect 8 on my computer,
without removing the approximately 100 megabytes of WordPerfect 8
program files. Because WordPerfect 10 offered only one feature
that was not found in WordPerfect 8 (the ability to save a document
in Adobe's PDF format) that I could conceive of using,
I did an uninstall of WordPerfect 10 and
re-installed WordPerfect 8.
After all of my experiences, WordPerfect 5.1+ for DOS is
still my favorite wordprocessor. My second favorite wordprocessor
is WordPerfect 3.5 for the Macintosh. Both are considered
thoroughly obsolete and are no longer supported by Corel-WordPerfect.
abandoning loyal customers
Many software products have an effective lifetime of only one or two years,
before it is discontinued by the manufacturer and replaced by a newer version.
This short lifetime is wasteful in several ways:
There is a widespread attitude in the computer industry, both hardware
and software, of abandoning loyal customers. Let me give a few personal
examples of such abandonment, in addition to Techwriter
- The old product often worked satisfactorily for a user,
so there was no rational reason to replace it with a "better" version.
- The "better" version generally used much more memory
(to implement rarely used features)
and was slower than the old version,
which was not an improvement.
- The "better" version took time to install and learn to use,
since software designers can not resist the temptation to
assign new functions to old keystrokes. I am not against
learning new software, but often the only reason to upgrade
is to fix defects in the previous version(s). In such cases,
there is no reason to change the "look and feel" of the software-user
- Files produced by the old version often need fine tuning by hand to look
or work the same way in the new version of the software, again, burdening
the user with the need to spend time to again be able to do useful work.
Hewlett-Packard Language (HPL)
The Hewlett-Packard desktop computers from 1976 to 1984
used a nifty language, called "HPL". Most of the commands were three-letter
abbreviations. Numbers were stored as binary-coded-decimal format, so there
was no need for a distinction between integer and floating point numbers.
The language processor was stored in read-only memory (ROM), so interpreting
the commands was fast. The language included the commands found in high-level
languages like Algol and FORTRAN, together with some commands from assembly
language, to give users a powerful set of commands.
Commands were included for controlling plotters, such as setting a scale
in the user's units, drawing axes, etc. Of the several computer languages
that I have used, HPL was the easiest to write and debug programs.
One could stop the program, interrogate the computer for the value of any
variable(s), change some values, then restart the program at any line!
Largely because of my good experience with HPL, I purchased a Hewlett-Packard
model 9825 computer in 1981 and a Hewlett-Packard model 9836 computer in August 1984.
Then, in Jan 1985, Hewlett-Packard suddenly announced that it was abandoning
HPL. Poof! My investment of many thousands of hours in writing HPL programs
was now a waste, as there was no hardware upgrade path and the computer would
eventually wear out. There were more than twenty-eight thousand model 9825
computers sold, and each one probably had at least one user who also spent
thousands of hours writing HPL programs. Therefore, it is quite plausible
that Hewlett-Packard's decision to abandon HPL wasted more than
3 ×107 man-hours of programming.
The value of this wasted time far exceeds the cost of these expensive computers.
Hewlett-Packard scientific/engineering workstations
During the 1970s, Hewlett-Packard invented the personal computer,
which was marketed to scientists and engineers.
During the 1980s, Hewlett-Packard continued to develop and sell
high-performance desktop computers to the scientific and engineering community.
There were two features that distinguished Hewlett-Packard desktop computers
from the mainstream:
As mentioned above in the section on
market acceptance, such nonstandard features
seem to cause computers to disappear without an upgrade path.
- Hewlett-Packard desktop computers in the mid-1980s and late-1980s used
Motorola microprocessors (the same choice as Apple Macintosh computers),
not the Intel microprocessors in the IBM-PCs and numerous clones.
- Hewlett-Packard used a proprietary operating system and programming
languages that were not used by any other computer manufacturer.
For example, Hewlett-Packard's version of the BASIC language
(sometimes called "Rocky Mountain BASIC", because the Hewlett-Packard's
desktop computer operations were in Loveland, Colorado) had many
useful features not found in standard BASIC.
In the early 1990s, Hewlett-Packard discontinued selling its
nonstandard desktop computers. Instead, Hewlett-Packard suggested
that its former customers purchase a clone of an IBM-PC from Hewlett-Packard
(e.g., Hewlett-Packard's Vectra or Kayak models) and install an
IEEE-488 interface card in that machine.
There were numerous problems with Hewlett-Packard's suggested migration:
- The old hard disk drives used with H-P's engineering workstations
used an IEEE-488 interface and could not be read on any IBM-PC.
Thus users would need to put the old H-P workstation next to the
new PC, connect the two machines via an IEEE-488 cable,
read one program file on the H-P workstation,
"print" that program file to the IEEE-488 port on the workstation,
read the program file on the new PC, and then store the program file
on the new PC. This tedious process needed to be repeated for
each program on the H-P workstation.
- The Hewlett-Packard BASIC software for the IBM-PC and clones
was not entirely compatible with the BASIC software for
the H-P workstations, so the user would need to revise some lines
of each program.
I spent substantially more money for the Hewlett-Packard model
9836 computer in August 1984 than
for my 1985 Toyota Corolla automobile the following year.
When I wrote this paragraph in May 2001, my Toyota Corolla still runs fine
and I can easily get replacement parts for it; however, Hewlett-Packard
has not supported its model 9836 computer for the past ten years.
IBM Professional FORTRAN
In March 1986, I purchased an IBM-PC/XT with the DOS 3.1 operating system and
IBM Professional Fortran. I chose this version of FORTRAN because:
IBM Professional FORTRAN worked fine on my PC-XT with its 8086 microprocessor
and 8087 numeric processor. I received an upgrade to version 1.19, with a file
date of 24 Feb 1986, from the local IBM representative.
- It was compatible with FORTRAN on IBM mainframe computers that I had used
in the 1970s. I wanted to continue to use my previous work.
- The IMSL mathematical and statistical subroutine libraries were available
for this FORTRAN compiler.
- This FORTRAN compiler was backed by IBM, a company that had a
strong reputation for serving its customers.
IBM Professional FORTRAN failed to work on my next PC, which had
an Intel 80386 microprocessor and an Intel 80387 numeric processor:
the compiler displayed a message "PROFORT requires math coprocessor"
and terminated without producing an object code file.
This FORTRAN compiler was written in 1984-86, before the 80386 was released.
This is simply a bug in the allegedly upward compatibility of software
for all microprocessors in the Intel 8086, 80286, 80386, 80486, ...
series. The problem is in the Intel chips, which do not
properly assure software written for earlier Intel microprocessors
that the numeric coprocessor is indeed present.
But it would be easier to upgrade the software than fix the Intel chip!
(In Sep 1998, I discovered that IBM Professional FORTRAN
would function satisfactorily on a computer with an Intel Pentium microprocessor.)
So I called several technical support telephone numbers at IBM in 1996
to get an upgrade for IBM Professional FORTRAN.
To my dismay, no one at IBM knew anything about this product,
which I had purchased just 10 years earlier.
Can you imagine an automobile manufacturer not providing spare parts for
a 10 year old car?
Part of the problem is that this was not "genuine" IBM software.
It was written by the Ryan-McFarland Corporation, and a
version was distributed by IBM as "IBM Professional FORTRAN".
I vaguely recall that Ryan-McFarland ceased business sometime
in the late 1980s or early 1990s.
It was bad enough that I had forfeited my $ 600 purchase price in 1986
of the IBM Professional FORTRAN. But I had also purchased a paid-up license
for the entire IMSL Mathematical and Statistical Libraries in object code for compiling
with my programs in IBM Professional FORTRAN, at a cost of $ 1200 in 1986.
This IMSL software was useless without the IBM Professional FORTRAN compiler.
So I purchased a new FORTRAN compiler with IMSL Libraries from Absoft in
June 1996 for US$ 1200.
Because I could not believe that
IBM Professional FORTRAN had simply vanished from the planet,
I made more telephone calls in January 1998 to various people at IBM,
but again no one had heard of that product.
AST Cupid Architecture
In 1991, I purchased an AST Premium computer with the microprocessor
mounted on a daughter board. The idea was that one could upgrade
the computer by simply replacing the daughter board, without also
replacing the case, power supply, disk drives, floppy drive, serial
and parallel interfaces, etc.
The AST catalogue in 1991 promised:
I purchased the AST Premium computer with an Intel
80386 microprocessor and I added an 80387 numeric coprocessor.
AST manufactured daughter cards with a 80486 microprocessor running at
33 MHz, and later daughter cards with a Pentium
microprocessor running at 90 MHz. That was the last model of daughter
card available for this machine. In 1995, when I wanted to purchase
a daughter card with a Pentium running at 133 MHz,
I was told that I would need to purchase a new computer,
exactly what AST had promised me only four years earlier that I would
not need to do!
- In a rapidly evolving industry, you need to be protected against
technological obsolescence. The solution is Cupid-32 technology.
- Simply put, Cupid-32 technology allows you to upgrade your AST ...
computer to state-of-the-art technology by simply replacing one
board with another in a matter of minutes.
- The savings both in terms of time and money are tremendous. ....
- Secondly, upgradeability allows you to extend the life of your initial
computer investment. Because the same system board, processor
board design and memory scheme are used in every Cupid-32 system,
there's a uniformity that ensures a high level of compatibility.
No matter how many times you change your processor board, system board,
It was a shame, because the full tower case was the best that I had seen in a
personal computer: it was easier to work inside than any other PC
that I have owned. The quality of this case did not come cheap:
the bare computer (i.e., a '386 microprocessor running at 33 MHz,
8 Mbytes of RAM, no graphics board, and no hard disk drive)
cost $ 4250 in June 1991, and that price was at a discount
from the manufacturer's list price.
There were ten EISA slots for cards, one of which
was needed for the daughter card with the microprocessor and another was
needed for a graphics card, leaving eight empty slots for expansion.
In comparison, most PCs in 1997 typically have three PCI slots and
only three ISA slots. I quickly use all three ISA slots for (1) an internal
modem, (2) second parallel port, and (3) a sound card.
I would like to add more cards to a modern PC, such as an IEEE-488 interface.
abandonment of DOS
The standard operating system on the IBM PC's and clones in the 1980s
was DOS, which was produced by Microsoft for IBM and other PC
manufacturers. DOS was the product that began the immense
fortune of Bill Gates.
In May 1994, Microsoft released DOS version 6.22, which was the final version
of stand-alone DOS.
When Microsoft stopped producing improved versions of stand-alone DOS, nearly
all application software developers also stopped producing new versions
of software for DOS. People who were happy users of DOS were thus frozen
in time in the early 1990s: if they wanted features of more modern software
(e.g., ability to export HTML from a word processor), they needed to purchase
a new version of their software that was designed for another operating
system (e.g., Windows 95 or Windows NT).
Many people dismiss DOS as a primitive operating system for dinosaurs.
However, DOS is faster and much more stable/reliable than the Windows 3.X
and Windows 95 operating systems. I happen to prefer the command
line interface in DOS to the menu interface in Windows.
The 640 Kbyte memory limitation of DOS is really not a big problem:
most applications software for DOS supports expanded, sometimes also extended,
memory. I routinely edit word processing files with sizes of 2000 Kbytes
in WordPerfect 5.1+ for DOS: the program is fast and reliable
even with these large files.
Corel, the owner of WordPerfect, was one of the last software developers
to continue to release versions of software for DOS.
Corel recognized that a large number of users prefer the stability,
reliability, and speed of the DOS environment.
However, in 1999, Corel discontinued sales for all DOS versions of WordPerfect.
Microsoft released version 7.1 of DOS as part of the Microsoft Windows98
operating system. However, Microsoft has no printed documentation for
this version of DOS.
The command-line interface in Microsoft WindowsNT 4.0 is a version
of DOS, but this version of DOS is not compatible with all programs
that were written for DOS 6.22 or earlier versions of DOS.
Fortunately, IBM has not abandoned stand-alone DOS users. IBM not only offers
that is year 2000 compliant, but also they promise "defect support
through 31 Jan 2001."
Similarly, Microsoft abandoned the large number of customers who were
using Windows 3.X. The last version of this operating system was released
in Nov 1993.
By 1998, it was difficult to find software developers who
were producing new versions of their software for Windows 3.X operating
systems, despite a large number of people who continued using this operating
system. For example, the Netscape version 4.08 was the
last version of Netscape browser software for Windows 3.X;
Netscape 4.08 was issued in Nov 1998.
Of even greater concern is that data files stored on current media
may not be retrievable in the near future. For example:
- The 5¼inch floppy disk was standard in the IBM-PC
(and nearly all clones) in the mid-1980s that used DOS or Windows 3.X,
but such diskette drives are not
supported in Windows95 and later Microsoft operating systems.
Mechanical parts in diskette drives and electronic components do not last
forever: when old computers fail, users will be unable
to read their 5¼ inch floppy diskettes on new machines with
modern operating systems. I stopped recording to 5¼inch
floppy disks in 1991, but I still have 600 of these old floppies
with archival backups of my wordprocessing documents and
distribution copies of software that I occasionally need to read.
- Apple included a 3½ inch floppy drive with their Macintosh computers
from the first model in 1984, up to about 1998.
Their iMac computers that were introduced in August 1998
never had a floppy drive. There was no floppy drive on the
Blue & White PowerMac G3 in a minitower case
that was introduced in January 1999,
or on the subsequent PowerMac G4 computers.
Apple users who wanted a floppy drive could install an external
floppy drive that was connected to the USB port. Such a floppy
drive could read/write to 1.4 MB floppy disks in Apple format
or either 720K or 1.44MB floppy disks in DOS format,
but not the 800K floppy disks in Apple format.
The floppy drives on the USB port can only read/write
floppy diskettes that are recorded with modified frequency modulation (MFM),
which uses a constant motor speed. The floppy disks formatted
for 800K bytes in Apple's older operating systems (e.g., System 7),
used group code recording (GCR), which uses a variable motor speed.
website for more information.
Apple simply abandoned the 800K floppy format in the late 1990s,
although Adobe had distributed expensive PostScript fonts on such floppies
at least as late as February 1995.
Apple users who have software on 800K floppy diskettes need to find
an Apple computer from the years 1992 to 1996 with a so-called
"Super [floppy] Drive" and make a copy of each 800K floppy
to the hard disk, then make a copy from the hard disk to
either a 1.4 MB floppy or CD-R disk.
By February 2002, 1.4 MB floppy diskettes were
scarce in office supply stores. Everyone seemed to be using CD-R
and CD-RW disks instead of floppies. It is not difficult to understand why.
In February 2002, quality floppy diskettes were selling for $0.40 each
(i.e., $0.30/megabyte), while quality CD-R disks were selling for
$0.70 each (i.e., $0.001/megabyte).
Hence, software that requires more than one floppy disk is
cheaper to distribute on CD-R.
I see a computer only as a tool for solving problems and doing
useful work. When I lose my investment in writing programs and
creating wordprocessing documents, I become angry. I do not like
wasting my time to repeat my work, just because some software or hardware
manufacturer has decided not to support an old product that I purchased.
I emphasize that I am neither against progress nor opposed
to learning new things.
But when my existing tools are adequate for my work, I resist
wasting my time to improve these adequate tools: I prefer to use my time
to do genuinely novel work.
My experience with the
Techwriter wordprocessor, the
and later, the Apple computer taught
me to avoid companies with a trivial market share. On the other hand,
such companies often have innovative products that are superior to
the market leaders, which is exactly why I chose to purchase them.
Developers of computer software assume that their customers
will continue to purchase every new version of their software,
which means upgrading software about once each year.
Because of this assumption, software manufacturers commonly discontinue
support for old versions of their product.
Worse, users of old software have data files (e.g., wordprocessing documents
or computer programs) that generally need individual hand-tuning
to function properly with current software.
Replacement parts are readily available for 15 or 20 year old
automobiles. Computer hardware and software manufacturers also need
to support their old products.
I do not expect hardware and software vendors to provide free
support for their old products. I am perfectly willing to pay $ 100
for an upgrade to software that neither (1) changes the "look and feel" of the
user interface nor (2) misinterprets data or document files created by
previous versions of the software.
this document is at
created 8 Jan 1998, modified 11 Mar 2002
return to my homepage