GridPad 1900: The first mobile pen computer

October 28, 2018

GridPad1900.jpg

As I wrote in my entry for the GRiD Convertible, GRiD was legendary computer manufacturer that produced a lot of “firsts”.

This one is (by and large) the first mobile pen computer. It was released in 1989, two years before NCR released their NCR 3125. It is quite heavy (2 kg) and the pen is connected to the tablet via wire, but it was the first time a company had the vision to give users something like an electronic notepad. In order to do that it offered (restricted) handwriting recognition. The project that led to the GRiDPAD was developed by Jeff Hawkins who would later on found Palm, and then Handspring.

Software- and hardware-wise it was quite far away from what GO and Apple would have in their (later) devices. Instead of an ARM-class CPU, and a special operating system fully exploiting and supporting the possibilities of an electronic notebook, the GRiDPAD has a meager 8086 and MS-DOS (the latter is at least built in and does not need to boot). Apart from a few applications, pen support mainly means that you can fill out text fields with the pen. No harddisk is needed (or offered) as it uses up to 2 battery buffered RAM storage cards as mass memory (up to 2 MB in total).

It is said that 10’000 GRiDPADs were sold in 1990 (probably its most successful year). It was marketed as a niche product mainly towards users with bookkeeping needs.

A later model 1910 had a built in 20MB harddisk, 2MB RAM, a NEC V20 CPU, and a backlit screen. The price for the 1910 was initially $3750.

The (probably) last model of the series was the GRiDPAD SL in 1993 that weighted 2.5 kg and costed initially $4395. This model could also run GO’s PenPoint operating system.

Technical Data

  • Manufacturer: GRiD
  • Model: GRiDPAD Model 1900
  • CPU: 80C86@10MHz
  • RAM: 1 MB
  • ROM: 256 kB
  • OS: MS-DOS 3.3 (built in)
  • Size: 31.4 x 23.5 x 3.6cm
  • Weight: 2 Kg incl. battery
  • Pen: passive, connected by a wire to the case
  • Display: 10“ LCD black&white, 640×480 pixel
  • Interfaces: RS232C (9 pin), keyboard (5 pin), external bus
  • Released: 1989
  • Initial price: $2370
  • Options:
    • Modem (2400 bps, MNP level 5 protocol=)
    • Hard disk extension unit (about the same size as the tablet): 40 MB HDD, 3.5″ FDD

Links

Advertisements

HTC Dream – the first Android phone

October 28, 2018

HTCDream

The HTC Dream, also known as the T-Mobile G1, was the first Android phone on the market. It was released in September 2008.

According to Wikipedia, ” An early prototype had a close resemblance to a BlackBerry phone, with no touchscreen and a physical QWERTY keyboard, but the arrival of 2007’s Apple iPhone meant that Android “had to go back to the drawing board”. Google later changed its Android specification documents to state that “Touchscreens will be supported”, although “the Product was designed with the presence of discrete physical buttons as an assumption, therefore a touchscreen cannot completely replace physical buttons”. By 2008, both Nokia and BlackBerry announced touch-based smartphones to rival the iPhone 3G, and Android’s focus eventually switched to just touchscreens.”

Although probably done “after having gone to the drawing board”, it seems this phone still breathes the before-touchscreen era having not only a keyboard, but also dedicated “phone up” and “phone down” buttons, as well as a trackball(!).

Although a 2008 phone, you can already load the battery using (mini) USB so there is no need for a proprietary power supply.

So it is a historically important smartphone, but is it also a rare one? Wikipedia quotes “In April 2009, T-Mobile announced that it had sold over a million G1s in the United States, accounting for two thirds of the devices on its 3G network.” So, not really rare 🙂

VCFB 2018: Short Report

October 21, 2018

As you might know by now, I attended the VCFB 2018 in Berlin. I had an exhibit and a presentation. I was most kindly mentioned on a popular German news page. I met many nice people.

The location was held at and co-organised by the German Museum of Technology  which was enough space and resources to support this event in a succesful way. Speaking of success: 2600 visitors have been counted, even more than last year.

Things-to-see-and-hear-wise, it was equally interesting. All the presentations were recorded and are available. There were many interesting exhibits. To me, the most interesting was “Smalltalk, Unix, Plan 9” by Angelo Papenhoff.

DSC_0030.JPG

He showed off:

  • A Xerox Alto simulator (Contralto) showing a very early version of Smalltalk
  • Research Unix (Bell Labs) and their GUI for it(!)
  • Plan 9 and its 9front forks (they develop the system further despite the fact that there is noone “official” from the original project left)

He gave also a presentation on “The GUIs of Research Unix and Plan 9” (German).

For the next year, I offered to contribute to the organization of the event, so maybe next year there will be more information on this event in English.

My VCFB 2018 Exhibit: PDAs using Magic Cap

October 21, 2018

DSC_0027c.jpg

At the VCFB 2018 I had an exhibit called “PDAs using Magic Cap”. It showed off a Sony PIC 1000 and 2000, a Motorola Envoy 100, and an Icras/General Magic DataRover 840. As the focus of the VCFB this time was “Graphical User Interfaces”, I concentrated on the Magic Cap GUI a little bit. Except the Envoy, all devices were up and running (I still have no power supply for the Envoy).

Interest in the exhibit was ok, the biggest group of people said my favorite sentence (“I have never seen something like that”), some were enthusiastic about the comic strip quality of the Magic Cap GUI, and a few people always wanted to have such a device.

There was also one (German) article about the VCFB 2018 which featured my exhibit and the presentation quite prominently.

If you want to read the posters next to the exhibit, either refer to this older blog entry of mine (English) or find it here (German).

My VCFB 2018 Presentation on “Mobile Agents and Telescript”

October 21, 2018

I had a presentation at the VCFB 2018 in Berlin on two topics on one of which I actually am an expert in :-). The title was “Mobile Agents and Telescript” and it dealt with the third topic of the General Magic topic: What would have happened technically if the first wave of Magic Cap devices would have been successful?

I gave my presentation in English for the sole purpose that you can also watch it as it was recorded (as all talks) by a CCC crew. So, without further ado, here is the link to the video recording of the presentation. And here are the slides that I presented that I presented.

VCFB 2018: I’ll be there

October 11, 2018

I’ll be at the Vintage Computing Festival Berlin (VCFB) 2018 exhibiting some General Magic Magic Cap devices and holding a presentation on Mobile Agents and Telescript. The VCFB start October, 13th and 14th in Berlin, Germany.

I would be thrilled to meet some of you there (I never really met a reader except people that read this blog because they know me).

I tried to be able to show the General Magic documentation movie there, but to no avail. Would have been a nice thing, rounding up the General Magic, Magic Cap, Telescript trifecta. However, I’m sure I will see it one day.

Exelbasic + (Plus)

September 2, 2018

exelbasicplusSome time ago, I acquired an Exeltel VX. Like any other flavour of Exelvisions computers (the EXL100 and the other Exeltels) it has no built-in programming language (in case of the Exeltel VXs this would have been a lesser problem anyway as it wants to be a Videotex terminal). However, if you want to load anything from a cassette (or a floppy disk or a battery-buffered RAM) you need your Basic. Exelvision delivered one with every EXL100, the Exelbasic.

In order not to have to borrow an Exelbasic from a friend (hi, cobrakai!), I bought one from Ebay (which does not appear that often as one wants, especially for a sane price). What I ended up with was not an ordinary Exelbasic, but the improved Exelbasic + (Plus) which I did not even know existed. Instead of a measly 200 pages of (French) manual, you have a 300 page one (also French). I am happy!

References

Intergraph Clipper Workstations

July 30, 2018

Just as an addendum to the last post, here is a list of all the Intergraph Interpro workstation models with a Clipper CPU.

We start with the earliest models:

Model Year of Introduction CPU
32C 1986 C100
120 1988 C100
125 1989 C100
220 1988 C100
225 1989 C100
340 1988 C100
360 1988 C100

From about 1989, Intergraph introduces a systematic naming scheme. It is described as a comment in the MAME emulator code in the Reference section. It follows an ABCD format:

  • A: case type (2=desktop, 6=minicase)
  • B: CPU type (0=C300, 4=C4T, 6=C400?, 7/8/5 = C400I)
  • C: graphics type (0=none, 3/5=GT, 4=EDGE-1, 8 = EDGE-2)
  • D: usually 0, 6xxx systems have 5, 7 and 9 options (backplane type?)

Graphics type 3 is for GT graphics fitted to a 2xxx system, and type 5 when fitted to a 6xxx system.

This scheme allows you, in principle, to generate all possible model numbers. However, of course, the CPUs and the graphic options developed over time and older options were not necessarily offered later on.

These are the models that I could find in the literature:

Model Year of Introduction CPU
2020 1991 C300
2430 1992 C400
2730 1993 C400
2830 1994? C400
3050 1989 C300
3070 1989 C300
6040 1990 C300
6080 1990 C300
6240 1990 C300
6280 1990 C300
6450 1992 C400
6480 1992 C400
6750 1993 C400
6780 1993 C400
6850 1993 C400
6880 1993 C400

Now, if you read the above scheme carefully and compared it to the list above, you can see that the scheme is missing some values. First, there seems to be a short-lived “3” case type. Second, there is an unkown CPU type “2”, obviously also a C300. Third, two unknown graphics options “0” and “7”.

Finally, I assume that the CPU type “8” is a C400 that runs on 50 MHz instead of the 40 MHz of the “7” models.

References

Intergraph 2700 Series Workstation

July 29, 2018

I bought on Ebay a 1993 workstation from a manufacturer

  1. you might not think about immediately regarding classical workstations
  2. that still exists today

The manufacturer is Intergraph, a company that started in 1969 on engineering for government agencies. It became one of the leading CAD/CAM suppliers in the 1990s, and stopped making hardware in 2000. Today, Intergraph is a 4000 people company working in the Geographical Information Systems area.

The workstation, a “2700 Series” model, is interesting mainly from a CPU point of view. It uses the last model of the “Clipper” family of CPUs (and not an ordinary PC processor, SPARC, MIPS, PA-RISC, or POWER chip, these are for common people :-).

The following paragraphs contain mainly passages from the [Wikipedia entry]. Clipper CPUs started their live when Fairchild decided to develop a 32-bit RISC CPU that was available from 1986. The first model was called C100. In 1987 Schlumberger, who owned Fairchild from 1979 on, sold Fairchild to National Semiconductor in 1987. They also sold the Clipper technology to the biggest customer of this technology, Intergraph. National Semiconductor would probably not be interested in Clipper, anyway, because they had their own 32-bit CPU the NS320xx family (in fact, it was the first 32-bit general-purpose microprocessor). It was developed earlier, and it was not RISC (but it was also not commcercially successful even if the list of computers incorporating this CPU is impressively large).

So in 1988 Intergraph developed a new model of the Clipper family, the C300. In 1990, finally, the last model of the family, the C400 was released. It was the first processor to integrate both architectural techniques, superscalar and superpipelined operations. Superscalar instruction dispatch means executing more than one operation at a time. Superpipelined operation means that some of the stages of a pipeline are broken up into smaller sub-stages in order to increase the performance. As a consequence, in 1990 the C400 with its 33 MIPs and 9 MFLOPs compared favourably to its competitors, e.g. early SPARC or POWER processors.

The Clipper architecture used a simplified instruction set compared to earlier CISC architectures, but it did incorporate some more complicated instructions than were present in other contemporary RISC processors. These instructions were implemented in a so-called Macro Instruction ROM within the Clipper CPU. This scheme allowed the Clipper to have somewhat higher code density than other RISC CPUs.

Unlike many other microprocessors, the Clipper processors were sets of several distinct chips. The C400 consists of four basic units: an integer CPU, an FPU, an MMU, and a cache unit, but the MMU and cache unit were combined into one CAMMU chip. Therefore, the C400 consists of 3 separate chips. Funnily, the largest chip is the CAMMU.

The strength of Intergraph (short for Interactive Graphics) computers was always graphics. Not so much the 3D graphics for VR and animated movies, but the 2D and 3D graphics for CAD/CAM-like applications. Therefore, also their workstations came with good graphics capabilities. My workstation even has a 2-display graphics card where everythings seems to exist twice: 2 x 2 MB Video RAM, 2 RAMDACS (yes, this was the time when RAMDACs were separate chips).

Intergraph workstations came for a long time in one of three combinations and names:

  • InterPro: desktop form factor plus separate display
  • InterAct: tower form factor plus two displays integrated into a digitalisation desk
    interact
  • InterView: tower form factor for a nearby digitalisation desk

According to my favorite market resource, at least in Germany, 3 different models of the 2730 desktop workstation were sold:

  • InterPro 2730-112: Single 19″ display
  • InterPro 2730-122: Double display
  • InterPro 2730-352: Single 21″ display

So, what model do I probably have? Well, after reviewing the above models and knowing the configuration of my workstation (32 MB RAM, 426 MB HDD, double display card), it is probably a InterPro 2730-122, and a not upgraded one.

You have probably noticed that I do not have pictures from a running workstation. That is, because I do not have a monitor yet. One would hope that any 13W3 display could do, but probably I am way to optimistic about that.

Like Sun’s, also Intergraph had a Forth-based diagnostics tool accessible at the firmware level. As I am a big fan of Forth, I am curious to see what one can do with this one.

According to [Wikipedia], “Fairchild promoted the CLIX operating system, a version of UNIX System V, for use with the Clipper. Intergraph adopted CLIX for its Clipper-based systems and continued to develop it; this was the only operating system available for those systems.” Later on, Intergraph ported Windows NT to Clipper computers, and even demonstrated such systems. However, Intergraph cancelled this project before its release. However, there is a 3.5″ floppy disk in my disk drive and it’s labelled “NT 3.5.1 21/21”. I’m very curious what this means. As soon as I get a monitor, I will tell you…

The last Clipper-based computer was the 2830 model which (I assume) is a 2730 with a 50 MHz (instead of a 40 MHz) C400. In 1995, Intergraph gave up this line of computers and sold exclusively x86 systems with Windows NT or Solaris.

The outer design of the workstation is quite standard and boring:

front.jpg

back.jpg

However, the build quality and the industrial design of the machine is really good. If we remove the heavy gauge steel cover by removing three screws we see a nicely laid out interior:

top.jpg

The right third is the PSU- The PCB on the bottom is the graphics card. Here is it in more detail:

DSC00030.JPG

The top part of the interior is a cage with the HDD and the FDD. It is so well designed that you can swivel the entire cage up after removing a thumb screw:

swungup.jpg

And underneath you find the SIMM slots to the left and the CPU on the right:

DSC00038.JPG

Four slots filled with 8 MB of RAM each, and another 12 slots empty. The chip with the big heat spreader (the one with the hole) is the CAMMU, the other two are CPU and FPU. If you want to see how the chips look like underneath: la voila!

Technical Data

Manufacturer: Intergraph
Model: 2700 Series
Introduced in: 1993
Initial price: ???
CPU: Intergraph Clipper C4 (a.k.a. C400) @40 MHz
RAM: 32 MB (128 MB max)
HDD: 426 MB (externally max: 9.4 GB)
FDD: 3.5″, 1.44 MB
Graphics: “GT+ Graphics”
Colors: 256
OS: CLIX
Interfaces: RS232, parallel, Ethernet, SCSI, keyboard, mouse, plotter, screen (13W3)

References

Tektronix Smalltalk Workstations (4400 and 4300 Series)

July 20, 2018

Tektronix (also known as “Tek”) is very well known for their manufacturing test and measurement devices such as oscilloscopes, logic analyzers, and video and mobile test protocol equipment. Lesser known are their computer models. Today I want to dig into the Tektronix family of Unix workstations that support a special feature: the programming language Smalltalk.

I am a big fan of Smalltalk even if this language today plays hardly a significant role. Tektronix was heavily involved in the commercialisation of Smalltalk; in fact they provided the first commercial version of Smalltalk.


What is Smalltalk?

As Smalltalk is currently such an exotic language (it was quite a niche language even at its best times), we should probably have a short look on this programming language.

Smalltalk is an object-oriented, dynamically typed, reflective programming language. The typical implementation uses a small (hardware-dependent) Virtual Machine that interprets bytecode. The major part of Smalltalk is implemented as bytecode (in this sense it is very similar to Forth). The typical Smalltalk system includes the Graphical Development Environment (using its own GUI if necessary), the runtime environment, and the Smalltalk-to-bytecode compiler, all written in Smalltalk. Also, all sources exist live in the system, i.e. you can read them, and you can change them on the spot. As a result, Smalltalk is probably the easiest language if you intend to shoot yourself in your foot, e.g. by editing the compiler that compiles the compiler code 🙂 You typically save a Smalltalk system by making a copy of the entire state of the Virtual Machine, including code, variable contents, the lot. Therefore, if you shoot yourself in the foot, you simply reload the last copy of the working Smalltalk system.

screen

Picture of how Smalltalk typically looked like in the 80s

Smalltalk is also a language that takes object-orientation very literally. Conceptually, the statement “1 + 2” is interpreted by Smalltalk as: there is an object “1” that is sent a message consisting of two elements, a “+” procedure name and a parameter “2”. Lesser languages aim to translate object-oriented code into a procedural intermediary language (like C++ was originally a pre-processor for C), but that’s not the Smalltalk way (I know that people knowing modern Smalltalk will now complain and try to explain that nowadays Smalltalk is able to e.g. directly compile into objectcode, but as we will talk about Smalltalk in the 80s, my explanation comes very close to the reality back then).

Smalltalk was developed from 1969 by Xerox PARC. In 1980 Xerox PARC made an attempt to interest other companies in creating commercial versions of Smalltalk. One of the companies that answered to this call was Tektronix.


The first development generation (1980-1981) of Tektronix’ Smalltalk system was a single-board 68000 system connected first to a Tektronix 4025 raster graphics terminal, later on to a directly connected bit-mapped display. The Smalltalk base system was implemented using their own Pascal compiler (cross-compiled on DECsystem 10/20 computers).

magnolia-st

Smalltalk on the Magnolia Workstation

The second development generation (Late 1980 -1981) ran on the “Magnolia” workstation. The architecture of Magnolia was very much along the lines with what we today identify as an early workstation: 68000 CPU, bit-mapped display, Unix-like OS, C compiler, harddisk, mouse. As the performance of the first Smalltalk system (like every other Smalltalk implementation apart from the original Xerox ones) on the single-board system was dismal, Tektronix changed the execution mechanism completely and ended up in a well-performing system.

This success inspired Tektronix to come up with a Smalltalk workstation product, code named “Pegasus”. It ended up in the model 4404 which was 68010-based and was marketed as an “AI machine”, featuring both Smalltalk and Lisp (as an option). As Motorola was not shipping yet a 68k MMU, a discrete MMU had to be designed as virtual memory support was considered essential for supporting Smalltalk and Lisp.

The third generation Tektronix Smalltalk was developed in 1985-86 for the 68020-based 4405/4406 models. It first shipped with the 4405/4406 in January 1986.

The final major revision of Tektronix Smalltalk was done for the 4317 model that was sold from 1987.

Fortunately, the history of Smalltalk at Tektronix is extensively documented (as an entry I suggest the “Tektronix Smalltalk Document Archive”, see the references).

As I am such a big fan of Smalltalk, I really would like to own a Tektronix Smalltalk workstation (a currenty unfulfilled desire). So, if you know one that needs a new owner…

Later workstation models did not came with Smalltalk, and Tektronix also made other workstation families. The Tektronix 6130 workstation (from 1984) had a NS32016 32-bit CPU, the XD88 family (from 1989) a Motorola 8800 CPU.

Please find in the following a list of the technical data of all 4400 and 4300 series workstations. Because it was little additional effort, I listed also the data of some 4300 series models that did not have Smalltalk.

4400 Series Models (Smalltalk)

4404.jpg

4404
CPU: 68010@10 MHz + NS32081 FPU
RAM: 1 – 4 MB
Resolution: 640 x 480
CRT: 13″
HDD: 45 MB
FDD: 320 kB
Interfaces: RS232C, LAN, SCSI, centronics
OS: Unix-like OS
Initial date: 1984
Initial price: $14950

4405.jpg

4405
CPU: 68020@16 MHz + 68881 co-processor
RAM: 1 – 5 MB
Resolution: 640 x 480
CRT: 13″
HDD: 45 MB
FDD: 320 kB
Interfaces: RS232C, LAN, SCSI, centronics
OS: Unix-like OS
Initial date: 1985
Initial price: $14950

4406.jpg

4406
CPU: 68020@16 MHz + 68881 co-processor
RAM: 2 – 6 MB
Resolution: 1280 x 1024
CRT: 19″
HDD: 90 MB
FDD: 320 kB
Interfaces: RS232C, LAN, SCSI, centronics
OS: Unix-like OS
Initial date: 1985
Initial price: $23950

4300 Series Models

2D Terminals (no Smalltalk)

4224

This family was the workstation variant of the 4320 workstation family. The models differ only in the grahics resolution.
4224: Resolution: 1024 x 768
4225: Resolution: 1280 x 1024
Graphics Processor: 68020@16.0 MHz
Colors: 256
Graphics RAM: 4 – 8 MB
CRT: 16 or 19″, 60 Hz
Interfaces: 3 * RS232C, LAN (IEEE 802.3), 2 * RGB, SCSI, color printer
Initial date: 1987
Initial price: $12950-$37500

3D Terminals (no Smalltalk)

This family was the terminal variant of the 4230 workstation family. The models differ only in number of colors.
4235: colors: 16/256
4236: colors: 256
4237: colors: 4096/16.7M
Graphics Processor: 68020@16.7 MHz
Resolution: 1280 x 1024
Graphics RAM: 4 – 52 MB
CRT: 16 or 19″, 60 Hz
Interfaces: 3 * RS232C, LAN (IEEE 802.3), 2 * RGB, color printer
Initial date: 1987
Initial price: $23500-$37500

2D Workstations, entry level (Smalltalk)

4317.jpg

The family is the entry line of 4300 series workstations. Apart from not having too much dedicated graphics hardware (it seems), they have a built-in Smalltalk environment.
CPU: 68020@16.7 MHz + 68881 co-processor
HDD: 86 MB
FDD: 1.2 MB
Interfaces: RS232C, LAN, SCSI, centronics
OS: UTek (Tektronix Unix with X Windows)
Initial date: 1987
Initial price: $9995-$18950
4315
RAM: 5 – 13 MB
Resolution: 640 x 480
Colors: monochrome
CRT: 13″
4316
RAM: 4 – 12 MB
Resolution: 1376 x 1024
Colors: 16 grayscale levels
CRT: 19″
4317
RAM: 4 – 12 MB
Resolution: 1376 x 1024
Colors: 16/4096
CRT: 19″

2D Workstations, medium level (no Smalltalk)

This family was the workstation variant of the 4220 terminal family. The models differ only in the grahics resolution.
4324: Resolution: 1024 x 768
4325: Resolution: 1280 x 1024
CPU: 68020@20 MHz + 68881 co-processor
RAM: 4 – 12 MB
Graphics Processor: 68020@16.0 MHz
Colors: 256
Graphics RAM: 4 – 8 MB
CRT: 16 or 19″, 60 Hz
HDD: 86 MB
FDD: 1.2 MB
Interfaces: 2 * RS232C, LAN (IEEE 802.3), 2 * RGB, SCSI, color printer
OS: UTek (Tektronix Unix with X Windows)
Initial date: 1987
Initial price: $23500-$37500

3D Workstations (no Smalltalk)

4330

This family was the workstation variant of the 4230 terminal family. The models differ only in number of colors.
4335: colors: 16/256
4336: colors: 256
4337: colors: 4096/16.7M
CPU: 68020@20 MHz + 68881 co-processor
RAM: 4 – 12 MB
Graphics Processor: 68020@16.7 MHz
Resolution: 1280 x 1024
Graphics RAM: 4 – 52 MB
CRT: 16 or 19″, 60 Hz
HDD: 86 MB
FDD: 1.2 MB
Interfaces: 2 * RS232C, LAN (IEEE 802.3), 2 * RGB, SCSI, color printer
OS: UTek (Tektronix Unix with X Windows)
Initial date: 1987
Initial price: $37500-$51500

Applications Processor (no Smalltalk)

4301

4301
This device completed the terminals to their workstation variant. It therefore included the CPU and co-processor, the CPU RAM, the HDD, and the operating system.
Initial date: 1987
Initial price: $14950

References

Tektronix 4300 Series Catalog (1988)
http://www.bitsavers.org/pdf/tektronix/44xx/brochures/4300_Series_1988_Catalog.pdf

Tektronix Smalltalk Document Archive
http://www.wirfs-brock.com/allen/things/smalltalk-things/tektronix-smalltalk-document-archive

Roger D. Bates: MAGNOLIA – A Single User System. Design and Implementation Plan. September 1980.
http://www.wirfs-brock.com/allen/files/tek/1980-9-16-magnolia-plan.pdf

Paul McCullough, “Implementing the Smalltalk-80 System: The Tektronix Experience,” in Smalltalk.80: Bits of History, Words of Advice, Glenn Krasner, ed., pp. 59-78, Addison-Wesley, 1983.
http://sdmeta.gforge.inria.fr/FreeBooks/BitsOfHistory/BitsOfHistory.pdf