Why was the 1401 such a
Run Away Success?

You can tell the author of this (above) proposal is very literate ;-))

My response (suggestions eagerly considered) was/is:
Unfortunately, the world is rather complicated -
Arguably the first general purpose "modern" business computer was either

• the British Leo, or the more actively marketed Ferranti Mark 1
• or the American Univac (I)

A person must remember a low/high point in IBM's history -
The United States Census Bureau accepted a Univac (I) March 31, 1951 and the Prudential Insurance Company was signed up for a Univac. IBM considered both their private territory.
T.J. Watson Sr. finally saw the handwriting and "got into computers"!!
In 1953 the 701 was "unveiled" and the 650 was announced.

In 1957 Watson Jr. mandated that all new designs will be primarily solid-state circuitry. (Pugh (1995) p.230) "Solid State in '58", causing a scramble for magnetic core drivers and printer hammer drivers ;-))

1. A fast enough hard nosed low cost business machine,
   enabling reduced equipment and machine operator cost.
   See 1401 Price Target $2500/Month
   "Hard nosed" - hardware multiply, divide, ... were optional extras.
2. Promoted by a company with a great reputation for sales and service,
   and which was already providing your present business data processing machinery.
3. A very early ?first? all transistor small business computer for small size and high reliability
   "A Third Survey of Domestic Electronic Digital Computing Systems" Report No. 1115, March 1961 published by Ballistic Research Laboratories
4. A great "swing" machine to ease a card "tab" shop to "modern" data processing
   (see below about "swing" machine)
5. Fast, reliable, easy to use peripherals.
   1. The fast (800/min.) card reader had two sets of brushes for read checking.
      - It also had 3 selectable output pockets to separate different function cards.
   2. The fast (250/min.) card punch had read brushes to assure accurate punching.
      - It also had 3 selectable output pockets.
      - If an operator blundered, and put in an already punched deck, PUNCH ERROR reported.
6. The new technology 600 line/min. printer made professional looking print.
   1. A technology that was world class for at least 20 years
   2. The printer drivers sensed for correct drive current -
   3. The Edit command made printing money fields on checks easy.
7. It was also ideal for media conversion, card to tape, tape to printer
   (as an auxiliary for a IBM 709, 7070, 7090, ... )
8. The random access storage devices (IBM 1405 and 1311) extended its range of usefulness
9. And the world class reliability and service, very important for business, continued.
   A note on reliability - IBM leased its equipment and supported the leased equipment. High quality manufacture decreased the support costs, making quality manufacturing worth while. IBM's competitors sold equipment, so low initial cost was more important to them - and junk breaks.

1. A person familiar with wiring (programming) an "accounting machine" such as an IBM 402, ... 409 could quite quickly come up to speed on programming an IBM 1401
   1. Easy, one character operation codes, IE Add, Subtract, ...
   2. A decimal machine, no decimal to binary, binary to decimal conversion hassles
   3. No "word length" problems, see Fran Underwood - 1401 - Design Philosophy. (A 16 bit mini-computer needs special software to add numbers totaling more than about 30,000 - $300.00 ), a 32 bit computer 2147483648 - $21,474,836.48 )
      -- with an IBM 1401, you can totalize our "national debt" directly.
   4. Unit Record people were already familiar with "field lengths", how may columns on the card were dollar amount? That many wires from card columns to say which print positions on the ( say 402 ) printer. Only COBOL, not available on small and midrange machines, gave you this independence from "tricky" code.
   5. A minimum memory 1401 could run the SPS assembler.
2. A new program, "Report Program Generator" was created to "to replicate punched card processing on the IBM 1401". specification
3. Flexible scope for change
   1. If you want to replace up to about 5 accounting machines, 40x,
      get a card only minimum 1401
   2. If you want to also replace your sorters and collators,
      add more memory and 4 tape units
   3. If you need random access to 2,000,000 characters add an IBM 1311
      or access to 10,000,000 characters of data get an IBM 1405.

--------------------------
Now - how to compress the above into a palatable sentence ?? ;-)) Ideally 10 words or less ;-))

• Customer program -
  A previous design (made by IBM Europe) had used a plug-board. Plug-board interfaces are expensive to manufacture - many contacts, circuits to prevent sneak currents, ... . And the customer is saddled with buying and storing many plugboards and wires.
  Magnetic core memory was just becoming competitive in cost sensitive machines. Customer programs could be loaded and executed from core memory at a competitive price, with the added convenience that data could be read into and out of the same core storage.
  To help save cost, memory access was narrow - one character at a time.

• Data width - how wide is your data?
  Business machines deal with large numbers of characters
      - such as National Debt
     - customer name, customer street address
  and small quantities
     - numeric day of month, numeric month, ...
  To handle these easily, compactly, business machines of the era used
     - character addressing,
     - a method of telling end of data
  The 1401 used a bit in each character which when set, indicated end of data field. This bit is called a "word mark". This permitted more data in less (expensive) memory.

• Memory access is character (byte) wide - one character at a time.
  This saves much memory access hardware, and cost.
  This results in a character serial machine.

• Serial operation, inexpensive (slower but fast enough) data moves and arithmetic. and as the memory was accessed a character at a time (above)
  The machine was "serial", as an example add two multi-character numbers together one character at a time. This continues until a "Word Mark" is detected in one of the numbers. Very simple, very inexpensive, and fast enough !

• How fast should memory be to satisfy the requirement to compute payroll and drive the printer at 600 lines/minute? (Faster is more expensive.)
  "Someone" - probably Fran Underwood in studies of the cost/performance problem determined that an 11.5 microsecond read/alter/re-write time was "good enough" - and matched currently available IBM techniques.

• How large should memory be ?? Core memory was quite an expense factor.
  There are stories of Fran going to customer sites, and hand coding programs for the proposed machine - code named "SPACE" which became officially the 1401. to match customer requirements. Program storage, data storage, input/output storage all needed to be considered. The minimum memory sold was 1,400 characters. The main memory could be up to 4,000 characters, and a separate box brought the total possible to 16,000 characters.

• How fancy should the operator's console be?
  In some early machines, for example the UNIVAC 1, the operator's console was large and imposing, and expensive. A very minimal operator/maintenance console was designed. As the machine was serial, fewer switches and displays were required. But the user interface was informative - data paths were marked on the console aiding operators in case of exceptions.

After the 360, IBM did not design nor market another character serial, variable field length, decimal machine.

There was an optional 1401 emulation mode in several mid-range IBM 360 series models, such as the Model 30, but the IBM and the world moved to fixed length, binary words for arithmetic. Characters were still individually addressable, but the word mark disappeared - forever.

Seemingly 99+ percent of the world's computing is word (several lengths) binary, in registers. There are specifications for decimal arithmetic, but these seem mostly academic, with no support in most compilers.

As remembered from Chuck Branscomb #1

On stage Chuck said the equivalent of what I firmly believe - Many people can build an electronic computer which will be moderately fast, moderately inexpensive, moderately reliable, moderately ... but to design and build peripherals which are reliable enough for business use is a challenge few companies have achieved. In an "off-line" conversation, Chuck Branscomb and I were discussing "Why IBM had such reliable and user friendly peripherals" I (with a background of very poor General Electric peripherals had proposed that reasons were: They were designed in up-state New York where farmers were practical and reliability was needed - ( a broken machine in planting or harvest season is a very serious, expensive problem ) . IBM leased the machines instead of selling them. Part of the lease was continuing service and maintenance - and a weak machine is costly for IBM to maintain. Other vendors sold their machines, and maintenance was likely another cost center. Chuck especially agreeded with the 2nd part, and added that IBM had long experience. For instance, Chuck had been in the Korean war (June 1950 � July 1953) involved with an Army "Data Base". Morning reports would come in from the many units reporting the presense and "functionality" of each of 30,000 ? soldiers. These had to be key punched, processed, and presented to "management" the following day. Reliable machines were highly appreciated/demanded. He remembered these lessons through out his career - including project managing the development of the IBM 1401.

As remembered from Chuck Branscomb #2

In an "off-line" conversation, Chuck Branscomb related this story about Fred Brooks and T.J.Watson Jr (IBM CEO). Both the hardware and software of the IBM 360 series was later than originally projected. The software was MUCH later, threatening to have 360 hardware ready for lease - but with no software. Fred Brooks had been the original software manager, but had left as scheduled to go back to the University of North Carolina to start a Software Engineering Department - which he did. The software development slipped further and further behind. Eventually T.J. Watson called Fred Brooks for a lunch meeting. Fred traveled from the university to the meeting. Watson spelled out the very serious software situation to Fred and asked for comment. Fred could of course have blamed the current project management - but didn't. Instead, Fred said the problem was his fault - a decision that he had made. Fred said that there were two software groups, the architects and the implementers. The architects said "don't start implementing until we are done architecting" and the implementers said that "If we don't start implementing now - you will miss your schedule." So I (Fred) went with the implementers, and cost you $100 million dollars. T.J. Watson's reaction was calm, and Fred went back to the University, - and wrote "Mythical Man Month". a classic recommended for all software people. (Robert Garner's comment - "Chuck had told me that Fred Brooks story too, but it's not clear about when Fred started his book. I'll ask Fred when I talk to him next. (So we can publish his 1401 story) )

As remembered from Robert Garner

On stage Robert Garner presented a result of the popularity of the 1401 About the time the IBM 360 was being developed, Honeywell, one of IBM's (small) competitors posed at threat to the IBM 1401. (IBM 1401 code could be run on faster, cheaper, (and good enough) Honeywell 200 machines.) Originally the IBM engineers said "impossible" then ??? watched a demo of a customer program deck (actually run through Liberator) by Honeywell and was impressed/very-worried, and reported back. Others reported that IBM lost an estimated 170 1401 sales to the Honeywell 200 - Robert Garner pointed out that the original 360 specification said nothing about emulating the 1401. IBM sales was very worried that the thousands of 1401 customers would have no clear upgrade path in the 360, and a high percentage would consider an inexpensive alternative, which IBM didn't offer. The IBM 360 mod 30 (with 1401 emulation option) and the later IBM 360/mod 25 could emulate the 1401 (but with out reader/punch pocket select). ("Fortunately" the IBM 360 was sufficiently complex as to make microcode useful in low and mid range machines.) Robert said that Fred Brooks told Robert that the happiest day of Fred's life was finding there was a micro-code solution so that the 360 mod 30 could emulate the 1401 and not risk all those 1401 customers - give them a viable up-grade path without having to re-code "everything".