||This article possibly contains original research. (July 2008)|
Foonly was a short-lived American computer company formed by Dave Poole, one of the principal Super Foonly designers as well as one of hackerdom's more colourful personalities. The company produced a series of DEC PDP-10 compatible computers, first the high-performance F-1, and later a series of smaller and less expensive designs. The first Foonly machine, the F-1, was the computational engine used to create some of the graphics in the 1982 film Tron.
The PDP-10 successor was to have been built by the Super Foonly project at the Stanford Artificial Intelligence Laboratory (SAIL) along with a new operating system. The intention was to leapfrog from the old DEC timesharing system which SAIL was then running to a new generation, bypassing TENEX – at that time the ARPANET standard. The F-1 was the fastest PDP-10 architecture machine ever built, with a clock rate of 90-100 ns per cycle, but only one was ever made. ARPA funding for both the Super Foonly and the new operating system was cut in 1974. The design for Foonly contributed greatly to the design of the PDP-10 model KL10.
The following few paragraphs are a personal account of the events, by Dave Dyer:
[Dave Poole, Phil Petit, and Jack Holloway came to Information International (Triple-I or III)] with a proposal to build an updated version of the original design (using ECL instead of TTL). I'm not quite sure how it came about - pretty crazy idea - but the connections between Triple-i and SAIL were deep and wide in those days. Triple-i was using PDP-10s for OCR, and for their groundbreaking movie group under Gary Demos and John Whitney, Jr. Triple-I had the usual grandiose plans requiring bigger and better computers. The three foonly principals spent about a year designing, constructing, and debugging the F-1. Poole was the mainstay, Petit was around quite a bit, and Holloway appeared only at crucial moments. My impression was that Triple-i paid the costs of construction and very little more - an incredible deal for Triple-i, considering that the F-1 actually worked. It would have been a very expensive boat anchor if it hadn't. I did a lot of work on the software - console computer program, a second version of the microcode assembler, and a port of TOPS-10 to run on foonly itself; and spent many fine hours with Poole, deducing I-Box bugs from errant program behavior. Shortly after the F-1 was operational, Triple-I and I parted ways and I mostly lost track of the F-1. Triple-i got out of the movie biz; the Foonly ended up following Gary Demos to several other early digital effects companies.
Foonly Inc carried on, building F2, F3, F4 and F5 in various quantities for people who wanted to own a PDP-10 but not to pay DEC's prices. The first few "little foonly" models were built from 2901 bit slice processors based on a design originally intended to be the F1's console computer. Alas, I don't think the F1 ever had a proper console - it always had some KA-10 attached. One of the first "little foonly" computers was sold to Symbolics for use as their original file server. Another customer was Tymshare Inc.
Foonly Inc. did not acquire any financial resources as a result of building the F-1. They turned to the market for low-end machines, producing a series of smaller, slower, and much less expensive DEC-10 clones that ran a TENEX variant called Foonex; this seriously limited their market. Also, the machines shipped as wire-wrapped engineering prototypes requiring individual attention from more than usually competent site personnel, and thus had significant reliability problems. Poole's legendary temper and unwillingness to suffer fools gladly did not help matters. By the time of DEC's Jupiter project cancellation in 1983, Foonly's proposal to build another F-1 was eclipsed by another DEC-10 clone, the Mars computer, and the company never quite recovered. Added by Phil Petit, (one of the above-mentioned Foonly designers):
The word "foonly" appeared one day as I was debugging my assembler, and typing in random nonsense to test the "ASCII" pseudo op. The word hung around, and got attached to one small project or another from time to time, until the project to build a new PDP-10 compatible machine came along. That seemed like a good thing to use the name Foonly for, so we did.
Many elements of our original Stanford design were incorporated into the KL-10 by DEC, with the permission of Stanford. In particular, almost the whole M-box (the memory interface and cache) was incorporated unchanged, except to replace TTL with ECL.
May 9, 2007[this quote needs a citation]
Added by Dan Martin - Principal Engineer for Tymshare Inc.
Tymshare sold the Foonly F4 to the Airforce for the Arpanet project. Several of the boards were converted to multi-wire process, and the console computer was replaced by an IBM PC running an application developed using Pascal. Interface boards were developed by myself that brought the total boot time from 15 mins to under 10 seconds. The old console computers were LISP derivatives that took forever to load. LISP was an artificial intelligence language that was not bad for a lab environment, but a nightmare for the market environment. I worked with Dave Poole for 4 years, and I have to admit he is very intelligent, was not a pleasant experience.
Sept 21, 2008[this quote needs a citation]
The main application for Tymshare's version of the F4 was a version of Doug Englebart's NLS system, developed when his team moved to Tymshare from SRI, called "Augment". The machine, called the 26KL, was marketed as the "Augment Engine" when running Augment.
- This article is partly based on the public domain Jargon File
- Tymshare 26KL Brochure at bitsavers.org
- Foonly documentation at bitsavers.org
- Lars Brinkhoff's table showing the F1 in perspective with other PDP-10 models
- Dave Sieg's notes and description of the F1