Return to A Short Biography

Next G.E. Computer Department Reunion to be late October, 2010 in Phoenix/Scottsdale.

Details in GE Computer Department Alumni Association Newsletter March 2009
2005 Mini GE-Reunion
A GE computer is found.

General Electric Computer Department
from the bottom up

1961 through 1965

by Ed Thelen

Work in progress
Updated through July 23, 2007 - Life on a G.E. Computer site Air Products, Trexlertown, PA

Warning - This monologue/tirade of G.E. sounds like a crotchety old man discussing archaic computing equipment. Well, I sounded the same when "IBM cards" were in and I was a new daddy instead of a grandpop.

In the glory days of computing - before the current standardization on the Intel 80486 architecture and Microsoft Windows - life was interesting. Any company that wanted a stock price surge announced it was getting into the computer business, hired a computer architect and some engineers that wanted to do something different, produced a computer that functioned, provided some EOM peripherals, sold a few systems, realized life was tough, and sold its computer operation for "pennies on the dollar". (And often the stock price jumped then also ;-))

Also - anyone who could spell the first four letters of "computer" could get a good paying interesting job ;-))

Now you have to be able to talk "tail recursion", "stack frame", "memory leak" and other odd topics :-((

- and if you can't program in the language of the month, C++, JAVA, PYTHON, ... , forget it.

---

General Electric Computer Department. - I was there January 1961 to January 1966

What promise!! - what happened??

From a high level perspective, I suggest:

GE COMPUTER DEPARTMENT - Event (Time) Line added Oct 2008, may help explain why this bootleg operation was always in money trouble, and therefore involved with such horrible peripherals. "King of the Seven Dwarfs: General Electric's Ambiguous Challenge to the Computer Industry" by Homer R. Oldfield - IEEE May,1996 ISBN: 0818673834 Here is a synopsis of the first 100 pages of the of the above book, and a discussion of the boot-leg creation of the "GE Computer Department". George E. Snively's corrections to King of the Seven Dwarfs. Note that George says that Homer ("Barney") “bootlegged” the GE computer business and was fired when GE’s Chairman, Ralph Cordiner – who had consistently rejected proposals for GE to enter the business - learned of it. A local backup copy Three articles in "IEEE Annals of the History of Computing" Volume 17, Number 4, Winter 1995. Articles are by J.A.N. Lee, H.R. Oldfield, and John Couleur. "My Adventures with Dwarfs: A Personal History in Mainframe Computers", by Russ McGee, released by the Charles Babbage Institute. (Several chapters of involvement with GE 600 line) on-line book in Adobe PDF format at: http://www.cbi.umn.edu/hostedpublications/index.html. Note that McGee's paper follows the lines of Oldfield's book rather than Snively's corrections :-| But says all occured before his arrival in Phoenix in Feb 1961 - about the time I arrived.

And I wish to contribute G.E. Marginalized by the IBM 360

However, I did not get to view the world from the top down,

I am a bottom-upper, and have a perspective as to

• why the G.E. field maintenance costs were so high.
  It is easy to argue that there was 1.5 field engineers per G.E. 225 computer site
  and 0.3 IBM Customer Engineers (CEs) per equivalent IBM 1401

  and why GE Computer Department had almost *NO* repeat customers!

  (G.E.'s only repeat customer seems to have been the U.S. Army at Huntsville,

  happy enough with their paper tape GE-225 system to order more.

  Their 225 was used as a high speed data logger for Saturn rocket development,

  not a business machine application.)

  A second customer did reorder, Canadian Imperial Bank of Commerce

Contents:

new engineer - off to G.E.

Phoenix & Computer School :-))

OK - what is a GE-225?

We're Off to Fix Computers ... :-))

Life on a G.E. Computer site Air Products, Trexlertown, PA

A Year at Elwood City a U.S.Steel town

And 2.5 in Phoenix HQ of GE Computer Department

G.E. 225 peripherals - listed in order of crappyness

Why the Krappy Peripherals?

And let's not forget "cost improvements"!!

Long Live Dilbert!! - How can Scott Adams get it (his cartoon strip) so right??

GE 225 vs. IBM 1401 probable main competition.

G.E. Marginalized by the IBM 360

Blame Game?? - who/what caused the failure of the GE Computer Department

Mail Bag

---

I'm a new engineer - off to G.E.

I graduated from Milwaukee School of Engineering at the end of the fall quarter 1960. I was hooked on computers having blown away part of my senior year on the school's LGP-30. I had already been in the Army doing Nike. At age 27, and in college and feeling time-a-wasting, had proposed to my girl friend - who started to cry. She said that she didn't love me but I figured two reasonable folks could work things out and we got married. :-)) and we now had one kid. :-)) After 24.3 years of my Dilbert type imperfections, and kids mostly out of high school, she left. Now 23 years later (2005), she is still searching. I figure it was a better marriage than most :-) but I sure was sad at the ending !!

I had previously worked at Honeywell's Aero Division, Minneapolis, as a gyro technician, for 6 months after the Army, and had an engineering job offer from them, but I wanted to work for Honeywell Computer Division. That division was interested, couldn't offer me a job because Aero Division had given me a job offer and corporate policy ... blaa blaa blaa

So I settled for my second choice, GE Computer Department, after talking with some folks at Johnson & Johnson of Racine Wisconsin who were maintaining the NCR-304 there. The NCR-304 was built under contract by GE Computer Department after GE had built the ERMA computers for Bank of America. The folks said that the water was fine, jump in :-))

Wife and I were almost dead broke at graduation. We had both worked to supplement the GI Bill. We had $100 from selling off much of our 2nd hand furniture at the end of college - and did not have enough money to rent a trailer and drive to Phoenix to the new job. I hated to ask my mother for the money - then a $200 graduation gift showed up from my Aunt Marion - and we (wife and 1st son Edward) rented a trailer and got from Milwaukee to Phoenix on that ;-)) (Gas was say $0.35/gallon and the tires were OK :-))

---

Phoenix & Computer School :-))

In Phoenix all I had to say was that I worked for G.E. and had instant credit everywhere. I even rented a house using the good name of G.E. - no deposit, no 1st month, no nothing. (This was before credit cards and that wonderful form of instant money.) As you can imagine, there wasn't much surplus left after the 1st pay check arrived. I was making $7,000 per year as a new engineer and in hog heaven! More than my father (an elected county official) ever made!

I had been hired to do field maintenance on the GE-210, the successor to the ERMA (above) but now G.E. was advising that joining the new GE-225 effort was a good thing to do. So I did. Our class of about 35 budding field engineers was given 6 months of very good schooling. I was in tune with the principle instructor - everything he said was just the right thing - at the right time - just like "good music". I soaked it up - delightedly. Everything he said seemed to resonate in my head. I had one other experience like that - Dr. Frier, a physics lecturer at the University of Minnesota. A wonderful experience. I had aced the good Doctor's physics classes and did well in GE-225 school.

Yes - you are listening to a guy who took about 7 years to get a Bachelor's Degree.

- Four years at the University of Minnesota in liberal arts,

- 6 months in a machine shop,

- 3 years in the Army with Nike Surface to Air Missiles,

- half a year at Minneapolis Honeywell Aero playing with gyros,

- and 3 years at the Milwaukee School of Engineering.
I am now about 75 years of age - and beginning to regard myself as "mature". - well maybe :-))

There was no Assembly Language assembler working when we started GE-225 class - we had to hand assemble our example and test programs ourselves. I was appalled - but was advised that a proper assembler would be available - real soon now - and in a month or so it did arrive.

The six months went quickly, learning CPU, core memory, the inter-unit protocol and hardware, Analex printer & GE Printer Controller, Ampex tape drives & GE Tape Controller, Elliott card reader - oh that sick puppy -, and were advised that the IBM 5?? card punch leased from IBM was so reliable that there was no need for maintenance or a maintenance contract (true!).

---

OK - what is a GE-225?

GE-225System Picture	For computer affectionatos, the machine was all transistors, except for 2D21 thyratrons in the tape units. The CPU had a 20 bit word, 8 K word (24 K character) memory with 21 microsecond cycle time with 3 core based index registers. (This is effectively 20,000 times slower than your PC and with about 10,000 times less memory. And about 4,000 times more expensive to buy in adjusted dollars.) You could buy a 16 K word memory. It, like most transistor computers was very reliable - maybe 1 transistor failed in two months. It was a competitive mid-speed computer at the time. The killers were the poor to horrible peripherals.
GE-235System Picture	Here is a GE-235 System Manual. A GE-235 was a code and peripheral compatible three times faster re-implementation of the 225 with faster, more compact logic.
	A real GE 225 exists :-))

A comment on a G.E 2xx architectural "detail", by Grant Saviers

My experience with a GE 215 or 205 (I forget the model number) was interesting and indicative of GE's future in the computer industry. A graduate course I was taking required a project on the EE dept. GE/EAI hybrid computer. A really ugly concoction of scaling and programming issues. I'll never forget the sign hanging on the GE console: "Please do not divide by negative integers as it hangs the computer so badly you will need to do a power on/off reset." It then listed the code in Fortran II to check for negative divisors and to get the right sign on the result. That code needed to be inserted before every divide operation. Fortunately, neither GE nor hybrid computing went much further. Grant

Baring the above described arrogance or ignorance or both,

I propose that almost anyone can make a functional computer. Well, OK - the really high end requires a genius like Seymour Cray. And OK, cost effectiveness and maintainability require some diligence. But usually, bright normal people assemble existing tinker-toys and get the collection to work at an acceptable speed.

*HOWEVER* I also propose that creating effective, reliable, easy to use peripherals is not so easy nor so common.

---

We're Off to Fix Computers ... :-))

OK - time to go to work :-)) Myself and two good friends were sent to G.E. Meter Department at Somersworth, New Hampshire to install and maintain their coming GE-225. Before leaving Phoenix, a personnel person handed out the expense forms and detailed their use. When we sent these forms back to Phoenix, we were told the forms were out of date and to fill out the new forms. Shades of Dilbert. !! How can Scott Adams get it so right?

Very pregnant wife and I rented a charming little white painted house in a rural setting overlooking a lake / farm pond. AH - YES, I forgot to tell you that for some years I seemly arraigned the family life so that there would be a major move just before/after birth of a new baby. So Carl was born as we were unloading the trucks and placing the equipment and connecting the cables and ... I wanted to be at the computer so that I would be aware, and not hopelessly left out of the computer scene. I drove the wife to the Dover, New Hampshire hospital then split for the computer installation. Heck, that was more interesting than standing around watch others chain smoke. For years there were family repercussions about my priorities !!

After a few months of really fun times, three good friend computer mechanics doing the work of one, (even better than the government - no office politics) the district manager said that I should go to Allentown, PA to talk with a customer about to get a new GE-225 and see how site preparations were coming.

So I took the family car to Air Products Inc. in Trexlertown, near Allentown (about 400 miles away) and met Carroll Claitor. Carroll wanted to know EVERYTHING, and I slowly realized that I was being interviewed. Apparently I passed. Returning to New Hampshire, I ran out of money, and arrived home smelly (no money for motel), hungry (no money for food), and only fumes in the gas tank. I did keep one dime for an emergency phone call if required. - again- this was before credit cards -

It slowly dawned on me that I would be leaving our happy little group of field engineers and going it alone. I told the district manager that I wanted to stay in Dover, but the district manager said that was not an option. Charlie Winters was going to stay in Dover, I was going to Allentown, Pa., and Jim ? was going to Norfolk. :-(( Another move - however, this time no late term pregnancy. - not enough time.

---

Life on a General Electric Computer Department G.E. 225 site

Air Products, Trexlertown, PA

This is being written in response to a request by Arvinder Singh Bawa, an Air Products employee, for images of GE-225 and Carroll Claitor, the manager of scientific computer at Air Products. I offer this little history of my minor part of the history of computing at Air Products, and Carroll Claitor.

- Emulating the LGP-30 computers being replaced

- Real world engineering computing

- How to play at work ;-))

- Remote diagnosis down to the slot

- Tit for Tat - Hot Stuff

- Hot Stuff, An Air Conditioning story

- Hot Stuff, Operating Margins, the techie view point

- Who cares, its Christmas Break Headquarters types

- Turf War vs IBM 1401 operation downstairs

- My impression of Carroll Claitor Manager of scientific computing

The G.E. 225 computer site at Air Products Inc. in Trexlertown, PA was "typical" or "average" - although the Auxiliary Arithmetic Unit was not common. Air Products ordered:

A G.E. 225 computer with 8 K 20 bit words of 21 microsecond memory - 3 connected refrigerator sized cabinets with operator's panel and print out typewriter. An Elliott Brothers card reader, 400 per minute An IBM 514 Card Punch, about 60 per minute An Auxiliary Arithmetic Unit - for hardware floating point - in 2 connected refrigerator sized cabinets 40 bit floating operations took 1 to 3 milliseconds. Four Ampex tape drives - density 200 bits/inch - state of the art at the time. A tape controller in a refrigerator size cabinet An Analex drum type printer - 900 lines/minute, 120 columns wide A printer controller in a refrigerator size cabinet A paper tape reader, not the one seen in the sales brochures A service contract which included Service Manuals, Schematic diagrams, test programs in deck form, and a plentiful supplies of spare computer cards, transistors, nuts&bolts, hand tools, soldering equipment, Tektronix oscilloscope, ... for the G.E. Field Engineer (me ;-) Backordered - a paper tape punch and controller. A GE computer is found. w/ pictures

Air Products had prepared a in a nice room with white raised floor and acoustic hanging ceiling. There were no glass walls for VIPs ;-)) The printer, printer controller, and card punch were in a side room also on raised floor with no door - for reduced noise for the operators.

We unloaded the cabinets and boxes from a padded furniture van and took them up the freight elevator to the second floor - the engineering area. (The first floor was for business, and had an IBM 1401 system for business purposes.)

A small room off the main room , with door, also on raised floor for convenience, contained the Field Engineer's supplies from G.E. (above) and work bench, various cabinets, shelves, ... The customer was responsible for supplying a desk and chair.

I bolted the cabinets together, connected the interconnects and cables, and in a few days the computing system passed all its diagnostic tests and was running customer assemblies and production runs. (The customer had access to some East Coast G.E. computer site during the order and installation period to enable a rapid start of productive programs.)

Emulating the LGP-30 computers being replaced

Part of the contract was that the GE-225 run, in emulation mode, the programs currently being run on the two LGP-30 computers now doing the inhouse scientific computing. About two days after the system was delivered to Air Products, a GE Applications Engineer arrived. Introducing Ruben Munz - business suit, cowboy boots, no hat, full of bounce, and ready to test his LGP-30 emulation in the real world. He would read in his emulation routine via the card reader, start reading the paper tape containing the LGP-30 program and data, and look for the result on the line printer. There was a persistent error on one tape. A particular paper tape frame was reading one bit incorrectly. The LGP-30 used pins to sense holes (10 characters per second) and we used a photo reader (much faster, maybe 60 characters per second) but the yellow paper tape was apparently translucent enough to cause this problem. We decided to leave the paper tape gain adjustments as specified, and had Air Products repunch the paper tapes into black paper tape stock. The emulation system now worked flawlessly and quickly. In 15 minutes, the GE-225 running emulation mode, with floating point subroutine calls trapped to go to the AAU, could do a full night's production of the two little LGP-30s. (I felt slightly bad about that because I had played too extensively with my college's LGP-30 in my senior year. It was a fun machine :-))

This customer, Carroll Claitor, readily accepted the fact that the machine was working as advertised - he was developing software and running production runs most of every day - and (after a visit by salesman Bill Peak) started paying rent or whatever.

With in a week, the two cute LGP-30s that our system replaced were gone to no one knew where.

Real world engineering computing

So - what did Air Products want a "scientific computer" for anyway? This computer was not just a prestige item to show customers, Air Products was a hard working, competitive (against the larger, well financed Linde Air Reduction) firm, wishing to grow and prosper. Air Products was in the business of producing compressed and liquefied gasses from air for medical and industrial purposes. (They also had an active helium program.) - Liquid nitrogen for cooling things, - liquid and compressed oxygen for burning things and iron, - argon, neon & krypton I presume, - and liquid helium, hydrogen and oxygen for the booming defense and space efforts. Air Products was in an expanding industry, knew it, and was eager for lower costs, sharper pricing, and more market share ;-)) Several computational things could help their business: designing more efficient, less costly heat exchangers for liquefying air designing efficient, economical fractionating columns (bubble towers) for separating the various air components "CPM" (Critical Path Method) to aid planning and ordering/construction scheduling of new Air Products factories to contain/utilize the above production tools - the GE-225 apparently had a good package as many customers went the GE data centers to run their problems. The walls in engineering were papered with paper of events in time sequence, when to order or start what, what had to be complete before which could begin that installation, ... Some of the diagrams were at least 30 feet long, with red lines (critical paths), green lines, changes, ... A useful version of "Linear Programming" using the newly developed "Simplex Method" of optimizing production was available. Inputting production and equipment and cost constraints, the program developed an optimal cost or production solution. The first two above were standard engineering challenges and there were FORTRAN packages and locally developed software to help optimize the designs. Items 3 and 4 (above) sold many GE machines to engineering customers. Due to memory constraints, 4 tape drives were mandatory for solution of practical sized problems.

How to play at work ;-))

So, this was my site. For the first time I was alone - in the Army Nike Ajax missile battery there were several technicians able to do what I was trained to do. Now I was solely responsible. If I did well, (superior up-time percentage) good things - I was delighted :-)) Fortunately the equipment was moderately reliable. The customer used the machine from about 7:30 AM towards 6 at night. There was no time clock on the machine - he could use it 24/7 if he wished, but he was paying for one shift Field Engineering coverage. Soon I was on the machine all night, programming, my all time favorite hobby, and was on call but asleep during the day. The customer was agreeable with this schedule as I lived about 10 minutes away and could be rolling in my car 5 minutes after the beginning of a call. A side benefit to the customer was that the equipment was well exercised and I retained/increased my knowledge of it continually. He also got free weekend coverage as I was around to play then also. I remember asking the customer if I could have a particular weekend off call to visit New York :-| If the peripheral equipment around the computer, such as the card reader and the tape units, had been reliable, life would have been even better. I could have played at work even more ;-)) But my techie life was good :-))

Remote diagnosis down to the slot

As part of my effort to provide superior up-time, I printed out a label for each card location, and scotch taped the label onto the card in that slot. This helped return the machine to a known state if someone got into swapping cards chasing a problem. So I was again made aware of the location of the adder in the 225. (Each card in the adder handled two bits and carries in that not dense discrete logic.) There were 10 cards to implement the 20 bit adder. About a week later, the customer supervisor called me at home saying "bit 13 in your adder does not work". (Did I say the staff was SHARP?). Like a fireman hearing the gong I was awake trying to remember where the components of that adder were. So I told the supervisor where to find what kind of card in my spares, and where to replace the card. In a few minutes he reported and all was just fine again, the new card worked just fine. Like a good Plains Indian Warrior, I had counted coup - had bragging rights ;-)) I also was SHARP !!

Others noticed that my site (Air Products) had a higher uptime than most. (This was honest up-time, not some negotiated deal.) And folks were aware that the computational chief - Carroll Claitor - was no push over - Soon my supervisor (Bob Kessler) heard that my tape diagnostics were finding weakness that the regular tape tests did not find. If your tape system passed my tape servo test, customer runs were not likely to fail from a misadjusted tape servo damper. I handed out other exercise routines that helped test the incoming reels of tape from GE and locate bad spots before they interfered with customer runs. (GE was getting really poorly manufactured tape, "warranting" that it would work, and customers were having a lot of trouble with the stuff.) I got to travel our district (eastern Pennsylvania) to help install sites and help if there was trouble.

Tit for Tat - Hot Stuff

One night I was playing at programming and the machine failed - it would not even run any standard diagnostic programs, but would pass all the little tests I would put in via the front panel switches. I was stumped !! About 2 AM I started to get seriously worried - I had no clue - no new ideas - this was not good - and the customer was coming in about 7:30 AM and expected a working machine !! I had a friend servicing the GE Missile and Space Site in King of Prussia, PA, just north of Philadelphia - about 50 miles away - I had helped install his machine. I called his home, waking him up, and I think I offered him my first born child if he would come right away and bail me out. He came anyway. About 3:30 AM he shows up and looks at my machine - and about five minutes later announced that the bit 8 carry in the program counter was not getting into bit 9. A quick card change fixed the problem, and back to home and bed he went. I was humiliated !!! How could I ever live this down?? Fate lends a hand - Not three weeks later, I get a call from my supervisor asking if it was convenient for me to got down to King of Prussia to help with a disabled Mag Tape Controller. The site was down hard - the customer HAD to have the tape system working - of course. So, hat in hand, ready to eat humble pie, I went 50 miles to help my King of Prussia friend - and would you believe? Just watching the tape controller lights I figured out the problem - in say 10 minutes - can you believe that?? Sometimes the fates are kind !! In any case, any time I think I'm HOT STUFF, I try to remember the above story.

Hot Stuff, An Air Conditioning story

Background: This generation of medium size and large size computers were (in general) cooled by air blown up through the circuit cards, ideally from below the raised floor - the computers ideally resting on a raised floor that had slightly pressurized cooling air. Other arrangements could be ordered, bottom mounted cooling fans and so forth. And why air conditioning? Primarily for the punched cards used for input and sometimes output. If the humidity was too high (wet) the cards were limp and tended to warp. If the humidity was too low (dry) the cards tended to warp another way. If a box of cards was highly humid, you aged or rested the card boxes for several hours/days in the new environment. This was especially important with the GE card reader made by Elliott Brothers of England. The cards leaving the read station flew edgewise through the air in the output hopper, hit the other end, and were expected to fall flatly down as the next card came flying by just above it. Any warpage or other problem and you had an output hopper jam with cards out of order, or you could get the cards out of order with out a jam :-| So - air conditioning was for: 1) the data/IBM/Hollerith cards, most important 2) cooling the circuit cards, *my* (germanium) transistors, and other equipment 3) cooling the operators so they didn't drip sweat onto the data cards. 4) cooling the room for any visiting VIPs 5) other, including employee/contractor/support comfort Now the Air Products GE Computer Room Air Conditioner story Air Products had a proper raised floor and good layout. The air conditioner and fan was on the roof, just above our second floor location, properly designed, installed and operating - for once "no sweat" ;-)) But after a few months, I started taking Saturday morning calls, my machine was broken. I would rush in - and find the room quite warm, and no air blowing through the cabinets to cool *my* transistors. !! For superior up-time, *MY* (germanium) transistors *WILL* stay cool!!! The first time no one could figure the access to the roof to turn on the air conditioner. Nothing to do but go home. The following Monday I/we got roof access figured out, and I put a sign on the switch to leave the GE Computer Room air conditioner on all the time. The next Saturday I took an identical call. I rushed in an up to the roof. (A friendly janitor had let me duplicate his key.) There was the air conditioning switch, sign removed, in the OFF position. So the following Monday Carrol Claitor wrote a note to all suspect groups that the GE Computer Room air conditioner was to be left running except in the case of emergency. And I taped on a larger RED sign on the switch. The following Saturday I took an identical call. I rushed in an up to the roof. That switch was OFF again, and my sign crumpled and discarded near by. - OK, - TWO can play this game!! I turned ON the switch, got my handy hack saw and sawed off the switch handle with the switch in the ON position. As you might guess, the next Monday there was a certain amount of tenseness - One maintenance group had complained that I had no right to saw off "their" switch handle!! The situation was resolved by moving the Air Conditioner switch to the entrance of the GE Computer Room, near the light switch. Turn it ON when you come in to start the computer, OFF if you turn off the machine to go home. :-)) - Building Maintenance was happy about the reduced electric cost, Engineering Computations or what ever was happy with more access to their computer, and so was I :-))

Hot Stuff, Operating Margins, the techie view point

Business types talk of Operating Margins as some difference between production cost and selling price. Techie types talk of Operating Margins as the difference between normal conditions and conditions of temperature, voltage, illumination, vibration, ... that cause a device to not perform properly. To help assure superior up-time, I wanted my Operating Margins to be a wide as practical. If the nominal air temperature was say 70 degrees, I wanted the machine to be fully functional over the normal air temperatures given by the air conditioner as it cycles ON and OFF. So, every month or so I would (at night) slowly lower the thermostat set-point until something failed, fix that something, then lower again, ... until I reach about 50 degrees. Then I would play the temperature game again on the up side until satisfactory performance at about 90 degrees. (There were thermistors near our core memory stacks to sense/cause modification of drive currents to cope with temperature changes, and they worked well :-)) Then I would play with the voltage regulators on the power supplies through say +- 10 percent. With plenty of Operating Margin, I was confident of helping provide Superior Up-Time ;-))

Who cares, its Christmas Break Headquarters types

The scientific computer staff at Air Products was really SHARP. One day I got a call that my Auxiliary Arithmetic Unit was malfunctioning. I raced in to find the supervisor (??Ray??) with a production run case that failed numerically. The supervisor had already entered in a test case into the machine with the front panel switches and the AAU clearly failed in the floating point multiply with these inputs. I ran the GE supplied diagnostic deck, with its test cases and there was no detected error. But clearly the customer's floating point numbers (with exponents near the limit of allowable range, one very large, one very small) did not multiply correctly. I added a test case card to the test deck with the customer parameters and the proper answer (in octal), and now this enhanced test also detect the problem. I swapped and checked the logic cards for weakness. Boards seemed to be working as designed. I checked the wiring in the exponent area, seemed to be proper, no missing or mislocated wires that I could find. I felt stuck - and did not want to get into the design of the AAU - I called my supervisor then Phoenix. (I called Phoenix a lot - ) Phoenix engineering would look at it. Several days and several phone calls later Phoenix said they could duplicate the symptoms, and there was indeed a design problem. The back ordered paper tape punch and logic had just arrived, and I wanted to get that up and running and on revenue quickly. And I really didn't want to reverse engineer that AAU - let the guys who designed it do that. The paper tape punch logic turned out to: - not well match the wiring in the chassis they sent me - be a work in progress, not complete, outputs to nowhere, inputs from nowhere. I spent my "Christmas Break" re-designing and reimplementing that controller. The unit and punch passed the diagnostics before the customer returned the first week of January :-)) Back to the faulty Auxiliary Arithmetic Unit - Then there was long wait - two week Christmas break for Phoenix. Of course I was getting pressure from Carroll Claitor - why shouldn't he call my AAU "DOWN" since it was capable of supplying bad answers - shades of the Pentium Problem floating point problem many years later - . My self and the salesman and the district manager waffled and weaved - we wanted to stay on revenue, the machine was working as designed, ... Our life was interesting - Carroll probably got some concession from the district manager. FINALLY - A few days after Phoenix came back from Christmas break, I was able to locate the ECO (Engineering Change Order) - in Phoenix. It was on some jerk's desk. It had arrived on his desk before Christmas break, but he saw no need to expedite anything - so he left it there over Christmas break - and what is the big hurry anyway - and why are you being rude? Anyway I got the guy with a bad case of headquartersitis to tell me the wiring changes over the phone with a promise to mail it to me ASAP. I put the wiring changes into the AAU and all was happiness - except my festering hatred of HeadQuarters types.

The above type of story was repeated over and over again - There was the tape unit relay problem - Oh - my head hurts -

All up and down the east coast, people were reporting occasional Mag Tape controller faults. Very long story - no clues - service people and customers were having trouble. Many tape runs were very long - say the sort of the records on a complete reel of tape - might take 6 hours. If in that time, you got an unrecoverable tape error or got a Mag Tape controller fault - you got to start all over again. You could spend all night starting all over again. - I figured that I lost 2 weeks of my life on this problem. Life was HELL. Eyeballing the scopes, where were these brief faults coming from that causes so much grief??

I found out later that a friend of mine, C.T. Winter, had taken spare logic cards, plugged them into unused slots in the Mag Tape controller, wired them up, and tried to trap the source of the problem. Whether this was involved with the solution I do not know - Wish I had thought of it.

Constant complaints to Phoenix - no help. Then we all got parcels in the mail - unannounced - new power relays to be inserted in the tape *drives* - the problem was that some contacts used to signal to the controller that the tape drive was powered up were power type contacts, designed to make/break high voltage and current. The replacement relays solved the problem by using wiping signal type contacts that were not so corrosion/vibration sensitive. Nobody told us that a solution to our nightmare had been found and that a fix was on the way - the complaining phone calls were handled by a different group than the fixing group - the left hand and the right hand.

Carroll Claitor stated on several occasions that he didn't understand how such a small organization (GE Computer had maybe 4,000 people) could be so bumbling and inept.

Turf War vs IBM 1401 operation downstairs

The GE-225 was working well enough. Since Air Products could use it 24/7 with no increase in price, there was plenty of free spare capacity just waiting. The IBM 1401 down stairs was metered, use it more than 8 hours/day and you paid for an extra shift. Carroll Claitor pressured management to let his department do some of the work normally performed by the business group down stairs. OK - try the ?weekly? invoicing run. The GE-225 had a reasonable COBOL compiler. Work began to run the IBM 1401 business tapes to generate the Invoicing printout. Soon we were doing short trial runs on normal paper then on samples of three part paper (with carbon paper in between). Our Analex printer did about as well on multipart paper as the IBM 1403 printer downstairs. The hammer impacts did smear the characters a bit, and cause slightly discolored rectangles of carbon - but that was state of the art - and satisfactory. The only real trouble was the way the Analex printer sensed "out of paper". We got around that by taping the four boxes of multipart forms end to end into one continuous form, four boxes long. The acid test - a live run, went well enough, and Carroll Claitor had a foot in the door of the 1401 operation downstairs. Then I went to Disk File school and transferred to a new site in Ellwood City, PA - and missed more Air Products excitement.

My impression of Carroll Claitor - Manager of Air Products ?scientific computing?

A truly memorable, quite unique character. Looked very similar to Barry Goldwater, Arizona Senator and presidential candidate. I don't remember him ever joking, he was unhurried serious. There was a very unique flavor of "we spin nothing here" aspect to him. He was the complete opposite of your used car salesman or (which politician to pick on - ) say Bill Clinton. You look at Bill Clinton, and you know he will hustle you, and you also know that he knows that you know, and he is going to do it anyway. There was a level of Buddha-like calmness in Carroll that was unusual, unique. Yes, the world is imperfect, yes there are problems, and yes we can analyze and deal with them - we shall overcome. I never heard Carrol tell a joke, tell anything approaching a falsehood - there was a calm, forward looking, no nonsense, inevitable pressure. The computing staff reflected Carroll's image - no jokes, no intentional distortions of reality, solid Germanic work ethic, due diligence, calm confident pressure forward. In the internal battles for turf in Air Products, Carroll had competitors who said interesting things. The thread was that some ?founder? of Air Products had rescued Carroll (an engineering graduate of ?Harvard?) from the literal gutter, brushed him off, and gave Carroll this present position. At the time I dismissed the rumor as silly/vicious - but how does one get a Buddha-like calm in our turbulent world? Has one gone through much worse? has known a bottom? survived? and is confident of better things? The only other person I have known that seemed to have that same quality of deep calm force was Helmut Mach, the Control Data salesman in Germany that developed a major computer order from Volkswagen. The word about Helmut was that he had been captured by the Soviets in the battle of Stalingrad, survived the horrible German/Soviet conditions, and as a captive of the Soviets until 10 years after the war had varnished/polished furniture in factories for the upper Communist Party members. Of the 250 thousand Germans captured at Stalingrad, about 13,000 (one in twenty) survived to eventually return to Germany. Helmut had that same calm, forward looking, no nonsense, inevitable pressure.

After many other adventures, and a year at a larger site in Ellwood City, PA I joined the inept headquartersitis folks in Phoenix. :-| I patiently searched for all the idiots I had to deal with from the field. They were all 8 to 5 types with all the fire of a drizzle. They couldn't care less about the products nor the well being of the poor fools in the field. I guess if you are a fizzle everywhere else, you migrate to "support the field". :-(((

Postscript:

1. After I left Air Products for the Ellwood City site, I heard that Carroll Claitor got the GE Philadelphia district manager to replace me with three people.
2. I understand that Carroll Claitor got an IBM 360 (? Mod 65 ??) in the mid 1960's rather than the GE 235 upgrade our district manager wanted to sell him. Maybe Carroll won his battle with Data Processing downstairs and took over more than the invoicing runs ;-))

G.E. Computer Department was trying to sneak into the general purpose computer business - and it showed!

1. The Elliott Card Reader/Shuffler

2. The tape drives were from Ampex.

3. Decca tape drives were imported from England.

4. Early Disk drives

5. Printers from Analex

6. High Speed Card Reader

1. The Elliott Card Reader/Shuffler - the shuffling was unintended !! 400 cards per minute

   The Elliot Card Reader/Shuffler

   (image from http://www.retrobeep.com/computers/elliott/elliottPeripherals.htm ) For the first two years, the only card reader option for G.E. 225 was the world's second worst card reader - from Elliott Brothers in England (Was there was a worse card reader? open to suggestions). After reading, the cards were ejected into the top of the output hopper (on the left) where they fell to the bottom for manual removal. If the cards were slightly warped you could get a shuffle, and/or a card jam, in the output hopper, and the cards would be out of order for the next operation or pass. There were other serious problems also - like filling the input hopper and emptying the output hopper and mechanical problems, but the above took the cake. G.E. added a thin piece of springy sheet metal about 3/4 inch x 6 inches to the output slot area, extending over into the output hopper on the left of the card reader, to try to deflect the cards down in good order. There was a little screw adjustment to position the springy metal to try increase its effectiveness. If you heard cards hitting each other or there was a card shuffle, you were advised to tweak the adjustment again. Different decks required different adjustments - And occasionally, the back end of the leading card would be hit by the front end of the following card, flipping the leading card over or causing an output hopper jam with cards in all orientations pressing up on the output hopper finger. To help keep mine going, I disassembled it every other weekend and repacked the bearing of the rollers that moved the cards with thick axle grease - which seemed to give sufficient damping of the card moving system for more reliable operation - like not getting mixed up on which column it was reading. - There were two methods of determining which column of the card was being read by the photo cells. The first method was used for the columns near the leading (left) edge of the card, while the card was being pushed by a little pusher arm and before it reached the output rollers. There were slots in the arm that indicated the position of the pusher arm, and hopefully the position of the card - but if there was not enough grease to damp things, the play in the pusher arm would whack the card and the card would bounce off the leading edge and an earlier column (or part of a column) was being read than indicated by the pusher arm slots. The second method was a chopper wheel on the shaft of an output roller. Card column determination was transferred to a counter of the chopper wheel. Dust could collect in the slotted aluminum disk chopper wheel and confuse the column counter - regular cleaning helped reduce this problem There was a check for end of card - if the end of card (all photocells lit) was too early or too late a card read alarm triggered. This check worked fairly well. You of course needed to make programs that could stop, let the operator take the card out of the output hopper, place it at the start of the remaining deck, and signal the program to re-read the last card. This machine was easy to despise!! OH - good grief - I forgot - the only safe way to reload the input hopper or remove cards was to depress the HALT switch on the front panel, then when done with those chores, depress the START button. But some people recommended lowering the REPEAT INSTRUCTION switch to do card manipulations, then raise the switch to contimue. This caused havoc to the results if there was an arithmetic instruction going at that time !! On the plus side, the Elliott card reader was evidently cheeep and not too bad if you only wanted to read one card. (Agreement) LaFarr Stuart claims the RCA card reader was worse - Who can tell? LaFarr had worked with IBM equipment and was hired into RCA as an expert. One day the RCA Computer Chief Engineer (with a background in communications) called LaFarr in and asked him what an acceptable error rate was for a card reader - how many errors per thousand cards. LaFarr said that no error was acceptable. The Chief Engineer thought LaFarr did not understand the question nor the statistics of the problem. LaFarr proposes the Chief Engineer did not understand the problem. You may contact LaFarr at lafarr via http://www.zyvra.org/.

2. The tape drives were from Ampex. Rewritten and extended again - July 2007

   Originally designed for light duty instrumentation recording service - they required a great deal of maintenance to keep serviceable in heavy duty business applications - such as sorts that can take hours of steady tape I/O - maybe 10 records/blocks per second - each record/block requiring a start and a stop - . One of the early symptoms that we were dealing with a light duty machine was the screw threads of the knobs that the operator twisted to mount a tape on the center spindle. After about 3 months of business use, with lots of tape mounting and unmounting, the threads wore and began stripping - making the tape drive unusable until and new nut in the center shaft, and new twist knob was installed. Soon a retrofit arrived, replacing the thread operation with a more robust levered handle. Lots of retrofits trying to convert a light duty easily aged drive to a more robust, low maintenance business data processing drive. Ampex didn't seem concerned that iron oxide is abrasive - the side tape guides were steel. After twenty to thirty hours of tape movement, the magnetic oxide on the magnetic tape cut grooves into the sides of the tape guides. The abraded groves shredded the sides of the tape. Retrofit #1 was to permit the sides of the steel tape guides to rotate, hopefully evening the wear - didn't work as advertised. Retrofit #2 replaced the sides of the tape guides with ceramic. That seemed to work for a while - but then long black threads of (I guess) melted/recooled mag tape began to plague folks - but I was leaving for a Headquarters software job - good luck folks!! Other quirks were, instead of tape vacuum columns (like IBM patented) these drives used vacuum "pucker pockets" to buffer the sudden starts and stops of the tape. The vacuum for the pucker pockets came from regular vacuum cleaner motors. The vacuum cleaner motors were *NOISEY*. There was no sound damping in the drive area - so being near four loaded tape units was like standing next to four vacuum cleaners - fatiguing!! Many people (including myself at both sites) bought extra vacuum cleaner hose and mounted the howling motors under the raised floor, which helped a lot. One fire marshal worried about motors under the raised floor, but I ignored him and he didn't press the point. Each pucker pocket had slot to permit the vacuum action to work. These slots were very noisey as the air rushed in to fill the vacuum. Fortunately they were mounted in front of the drive main plate so the sound was largely muffled by the transparent plastic door. The brushes on these motors lasted maybe 500 hours, then would start to arc destroying the commutator and causing the drive to fail - hard - like replace the vacuum motor. Of course you should add checking the length of the brushes to your monthly inspection - but there went another 15 minutes per tape drive. OH - YES :-| The tape drives were manufactured in temperate California, and shipped from temperate Arizona. The hardened glass of the pucker pockets was epoxied to steel hinges so the glass could be swung out for cleaning. Unfortunately, the steel and the glass have quite different temperature coefficients of expansion - and if you received a system in a northern winter, the glass pucker pocket covers were usually cracked, and you had to send back to Phoenix for replacements. If the weather stayed cold, the replacements also cracked in shipment. So some regional CE or salesman would fly to Phoenix to hand carry a supply back to the cold region - say Wisconsin or Pennsylvania. Oh - how can I forget - tape drive head alignment - As delivered from GE Phoenix, the heads were often misaligned so much that tape written on one drive could not be read on another drive. One of the things the installation crew would have to do is at least make all the drives on the site interoperable. Preferably, a customer tape from say an IBM system would be used as a reference if the GE system had to exchange tapes with other system. What slop!! Speaking of slop - The tape supplied by GE was so contaminated (1 mm white surface particles) that the field people were well advised to write the tape from end to end (ignoring errors) several times before the customer used it. That way the stray particles left over from the manufacturing process would be (mostly) scraped off by the retrofit tape cleaners (that looked like the shaver foils from an electric razor). *really* I imagine the only people happy with GE supplied tape were the GE bean counters - I would have mentioned the tape packer arm - to try to solve the "cinch" problem. The different wraps of tape on a reel would sometimes slide over each other, creating sharp wrinkles at 90 degrees to the side of the tape. This crinkle or "cinch" would set in the plastic of the tape and create a little line of tape that couldn't be read later. Adding to the seriousness of the problem, about one in 2^7th cinched records would pass the hardware checking, but not contain all the data originally written.) There was no hardware length check with this tape validity checking scheme. Years later I found that the IBM 729 tape drive we were competing with had the same problem. :-| Apparently this was a generic magnetic tape problem caused by torqing the reels. I had forgotten why we were so busy - Reginald W Oldershaw, who later worked at Ampex, mentioned the following Ampex oriented web pages - Ampex Virtual Museum and Mailing List - Preliminary Guide to the Ampex Corporation Records, ca. 1944-1999 Online Archive of California, Stanford

   Later (on the GE-625) Decca tape drives were imported from England. They used a "scramble bin" instead of the tape columns that IBM used, or pucker pockets that Ampex used.

   • But the tapes needed 25 foot leaders and trailers to properly fill the scramble bins -
     instead of the usual industry standard 8 foot leaders and trailers. (Interesting compatibility problems with other machines - even other GE machines. Customers had to patch on 20 feet of tape onto the leader and trailer of any standard tape they wanted to use on the G.E. Decca tape drives. How could G.E. engineering "management" be so $%^&*( stupid ??)

     If there was any fault, a *pulse* was emitted from any of a number of sources to a latch that lit the blue failure light and shut down operations. There was generally no way to determine where that fault pulse came from. C. T. Winter (RIP) tried to get peripheral engineering to make more user friendly electronics - Ha -

     The systems containing DECCA drives had much more than the usual tendency to stretch and break mag tape - I don't know if it was the English electronics or the added GE electronics. The GE hardware people claimed that the fault was the COBOL SORT, and convinced GE Phoenix management to have the software folks examine the sort generated by COBOL for situations which could cause commands to the tape controller to break tape - I kid you not!!

     (From the above point) Until GE Phoenix eventually hired an IBM guy (Haafstra), management had little clue about data processing. The GE mantra was that "a good manager can manage anything". Maybe if the product is as simple as a motor or a transformer or a light bulb - but in the complex conflicting arena of data process, GE managers were only bean counters with no computer industry knowledge to keep them from silly blunders.

   Eventually a new general manager correctly called the DECCA drives "boat anchors" and consigned them to the appropriate place. I have no clue about what drives replaced the Ampex and DECCA drives -

   Early Disk drives - "35 million decimal digits"

   - OK lots of folks had troubles, but I contend "we" likely had more trouble. As I was leaving the field for Headquarters, there was a team going from site to site with automobile polish to slick up the disk surfaces. Unfortunately, any extra polish tended to slowly build up on the flying heads, destroying their ability to fly. This cure was worse than the original problem. People running the current "Winchester" technology are totally spoiled! And of course, GE had to make a cost improvement that caused a lot of grief for a few months.

   Printers from Analex

   - the usual vertical dispersion problem rather than IBMs less obvious horizontal dispersion problem. Ah - yes - and the usual user hostile GE controller - If you ran out of paper, the printer would reset, destroying the data line stored in the printer. For various user unfriendly print form alignment problems, there was no way to stop of the printer at the end of a form. IBM went to the extra trouble of sensing out of paper *after* the last line on a form. The just completed form could be removed and the next form inserted just below the print hammers, and life was easy. GE and most others sensed out of paper about 6 inches below the print hammers and life was tough. Really tough when running say payroll, billing, etc. Where several boxes or more of expensive multipart forms were getting printed in one run - customers had to prepare the expected numbers of boxes of paper, tape the beginning and ends of the boxes together into a continuous stream, and start out. "Been there, done that"

   And the later High Speed Card Reader

   - This reader actually worked quite well - but was hell on cards. The cards were separated by an air-blast to get individually vacuum picked, transported by speeding belt past photo cells for reading, then stopped in the output section by impacting (smashing) them into a block of steel. The result was that the cards got limp rather than retaining their stiffness - and after about three passes through the High Speed Card Reader you were well advised to reproduce the deck with say an IBM 514 card reproducing punch. A deck going through the GE High Speed Card Reader say ten times was guaranteed to give jams of one nature or other. Years later I used a Control Data 504 High Speed Card Reader using the same principle - but with a block of plastic to stop the cards instead of a block of steel - and the cards lasted "forever". In "fairness", the card handling technique was very similar to the quite successful Pitney-Bowles document handler (reads checks at banks) that GE used, then manufactured. BUT, - the documents went through the machine only once per bank/distribution point - were much thinner and not supposed to be stiff - and were much lighter and I presume slowed a lot from friction in the output hoppers.

As I was leaving, G.E. tried to make/market a head contact disk, plated with rhodium which was to provide lubrication for the contacting head, that was supposed to be a break-through - well, it just cost G.E. another black eye, lost time, lost money, lost customers, ...

Customers came to hate many GE peripherals, and hardly ever selected GE computers more that once. As much as some folks hated IBM's high handed marketing and prices, at least their equipment worked well and was maintained well.

Customers were caught between the Devil (GE) and the Deep Blue Sea (IBM).

---

• The disk heads that flew - off
• The power supply bolts that smoked
• The core memories that *sometimes* forgot
• Make one prototype (G.E. 625) and time share it for hardware and software development

The GE Computer Department was basically a bootleg operation - GE Headquarters had issued an edict that GE would *not* get into the general purpose computer business, which was exactly what GE Computer Department leadership was trying to do !!

The results were that GE Computer Department did *NOT* have the deep pockets of the GE Corporation - although many employees assumed and acted as though the pockets were very deep indeed. (Another story for much later.)

In any case, GE Computer Department was always strapped for cash, and tried to make up for it by emphasizing "Cost Improvements". OK, how can such a good idea go wrong? Well, make the proposed "improvements" without checking with designers or knowledgeable people !!

False (very expensive) cost improvements were the rule, not exception. The slightest quick check with designers could have prevented many/most of the disasters. Unfortunately the "cost improvement" program was badly administered. Groups were rated by the number and estimated dollar saving of "cost improvements" they made. If the group checked with a designer or competent person, the "cost improvement" might be shot down before being counted - so there was a negative reward for checking. :-((

Where to start in this sad list?

• The disk heads that flew - off

  The big GE "hard drive" (# 4 above) had a brief, special trouble. New production units would work "reasonably" well for a while - "reasonably" for that era in history - then individual flying heads would break loose and reek horrible havoc - crash other heads, score disk media, ... Investigation revealed that the beryllium-copper flexible head supports were breaking - but that the beryllium had been cost improved out. The resulting copper shim stock fatigued and broke in a few months. (As any mechanically inclined person would have predicted.) The resulting retrofit costs and loss of customer confidence were severe.

  The power supply bolts that smoked

  Power supplies (like 50 pounds each in these good old daze) had bolt conduct the output power. You connected the power wiring (like *6 volts) and ground to these 1/4 inch diameter bolts.

  Suddenly, the power supply bolts of new production power supplies were getting very hot. The phenolic back panels that the current conducting bolts went through discolored and weakened and caused bad smells.

  Investigation revealed that the copper bolts that had been used to conduct the 20 or so amps of + 6 volts had been cost improved by substituting iron bolts from a hardware store instead. The iron, having much higher electrical resistance that copper, got hot and caused the problems.

  It was impractical to field disassemble the power supplies to replace those iron bolts and discolored phenolic back panels. New power supplies were shipped to the GE sites, and the local field engineers had more work to do - like life wasn't full enough already. Those Maytag TV ads of Maytag repair people with nothing to do were not GE field engineers!!

  The core memories that *sometimes* forgot

  Intermittent problems are the bane of any computer fixer's life. You suspect that a part is failing - but when you test it - it works just fine - until the customer uses it for a few days again.

  The DataNet-30 (a communications computer) was an 18 bit word machine, (18 expensive core planes) with out parity which would have needed another core plane and supporting circuits. However, GE core memories had been very reliable (the ERMA for the banking industry, and the 225 that I had helped maintain in the field) - so the parity checking was eliminated in this product.

  Except - new production DataNet-30s were having *A LOT* of trouble in the field.

  • - The hardware folks thought it was the flaky software folks and their flaky software - and customers were allowed/encouraged to program the real-time communication machines in assembly language.
  • - The software folks figured the complicated hardware was flaky.
  • - GE programmers and field engineers were flying all over trying to correct the problems and pacify irate customers that refused to pay for "their" DataNet-30s that were not doing the job.

  After a few months it was noted that memory diagnostics could run for days, then suddenly fail, then run just fine again. What could be causing this?

  The internal GE "Core House" had cost improved the soldering of the wires between core planes by not twisting the wires before soldering - the wires were just pinched together then soldered. In the field, these soldered joins could fatigue and crack just slightly. Conducting adequately sometimes, and not sometimes. Parity checking would have instantly identified the extended problem as a memory failure - and any soldering expert would have vetoed a pinched/not twisted solder connection.

  And the bean counters did not assign a cost to this whole fiasco either. - who could? - A very risky change ("anyone familiar with solder joints would veto the change) saved maybe a person-week and cost hundreds of thousands of dollars in transportation cash, payroll cash, replaced memories, lost revenue, lost referrals, increased "cost of sales", lost reputation, lost "good will".

  Make one prototype (G.E. 625) and time share it for hardware and software development

  Save money not making a second prototype - and lose it on increased engineering labor costs and increased time to market.

  For better or worse, the basic design of the G.E. 625 series was imported from G.E. Schenectady (where it had been developed for a military contract?) As explained in Homer R. Oldfield's book Software had the prototype computer for 12 hours/day, and hardware had it the rest of the time. But when hardware got off the machine, it might or might not be working. Or it might be working except some instruction might not give correct status or something - anything. The result was that software was badly delayed, hardware and software fought (what else is new?), and hardware development badly delayed also. Who can say how much that decision to save a second prototype cost in blown engineering salaries and late time to market? The bean counters seem to think other people (engineers) are free when making decisions!! I love Dilbert!!

  Its getting late - more, many more, enough to bore, later - the list seemed endless but I forgot lots.

---

In summary - the GE Computer Department "Cost Improvement Program" had many high profile failures - due (I contend) to management inattention and blunders. I cannot imagine that the successes out weighed the failures by a high ratio.

I imagine the bean counters counted the saved paper clips and re-used typewriter ribbons, and missed the more complicated, embarrassing costs - bean counters are paid to make management look good :-)) :-((

---

"Buy GE policy"

> Glad you like the noises!  ;-)
> 
> Robin thought the tape drive sound (the vacuum "fart" when the    
> tape loads)
> sounded like an ER surgical procedure!  ;-)
>