Talk:Punched card/Archive 2

Chad
Question about bits and chad. I worked with punch cards for years. All the places I worked refered to chads as bits (bits of paper), they were even collected in the 'bit bucket'. I never heard the term chad until the 2000 election! Why do so many sites claim bit is short for binary digit instead of 'bit of paper'? Is there any proof that either is correct or incorrect?

. Thanks Dwight DBArrants@starfishnet.com


 * See Chad (paper) and . --Zigger 04:01, 2004 Jul 4 (UTC)

The term always used at IBM was neither "bits" nor "chad" but rather "chips." The container which received the punched-out pieces of card was called the "chip box". The term "chad", I believe, was more commonly used in reference to the little paper dots punched out of paper tape. My "authority" on this derives from the fact that I was once employed by IBM as a Customer Engineer (service technician), and my father was once an engineer with the FAA, and worked for many years with teletypewriters (used for aviation weather reports). I played with TTY gear myself, together with my father, in connection with Ham Radio. Dad called the tape punch-outs "chad" and at IBM, the analogous part of a punched card was always a "chip."

I suspect that the references to "chad" in connection with the 2000 election problems in Florida may result from the fact that older news reporters are more familiar with the TTY term, as TTY equipment was used to receive wire-service newsfeeds, often relayed by means of paper tape. Some paper-tape punches were designed to punch little "trap doors" in the tape instead of removing the paper altogether. The holes could still be sensed with pins, but the text could also be printed on the tape because the holes weren't punched out. I have one such antique device (ca. 1942) in my garage, officially called a "Teletype Model 14 chadless printing re-perforator". --RussHolsclaw 06:14, 31 January 2006 (UTC)


 * My 1967 Random House Dictionary of the English Language defines chad as "the small paper disks formed when holes are punched in a punch card or paper tape."  --agr 15:51, 6 March 2006 (UTC)

"Disks" eh? (emphasis above added). Could it be that if the holes are round, they're chad, but they're chips if the holes are rectangular? :-) RussHolsclaw 05:09, 7 March 2006 (UTC)


 * There is a discussion of etymology in the article on chad (paper). It was not unusual for IBM to employ terminology that differed from the rest of the industry. --agr 11:53, 7 March 2006 (UTC)

You seem to be suggesting that I think chad is a newer term, while citing an etymology that states that the term originates in 1947 (which is wrong... it's much older than that). I don't know which term was the original one, I'm only stating that IBM always seemed to use the term chips, not only internally, but in customer documentation, and even on signs and indicator lights on the equipment. For example, the 2540 card reader/punch, and its predecessor, the 1402, each had an operator's light labelled "Chip Box", which lit when the box was full, or had not been put back in place. Interestingly, it determined both conditions by weight; too light, no chip box; too heavy, the box is full.

I inherited some old Teletype maintenance manuals from my father, dating back to 1942, although he worked on TTY gear even earlier than that. The book on the Model 14 tape reader ("Transmitter-distributor") shows a picture with chadless tape, and a parts book from 1944, for a tape punch unit ("perforator"), lists a "chad chute" as one of the parts. Both of these prove that the etymology giving 1947 is wrong, and probably by much more than the dates in my WW II-vintage manuals.

Still, IBM used the word "chips" for card punch-outs, and undoubtedly did for at least as long, according to the old-timers I met when I first started working for IBM in 1966. It seem probable that Herman Hollerith called them that, too, or IBM probably wouldn't have used that term. It would be interesting to see which term is older. Teletype (AT&T) used "chad", and IBM used "chips". I wonder what term was used at, say, Remington Rand, with their round-holed 90-column punched cards, or Frieden, with their Flexowriter paper-tape typewriter. You seem to be pretty certain that "the rest of the industry" was using the term "chad". Do you have evidence for this assertion?

[Friden (only one "e", often misspelled reasonably, as here) definitely did call it chad. I worked for Rochester R&D for a spell; that's where the Flexowriter was engineered. Regards, Nikevich (talk) 19:17, 10 March 2011 (UTC)]

Speaking for myself, I knew the term "chad" earlier, from my father and the TTY. I only inserted the reference to "chips" because it is the customary word to use in connection with IBM punched cards, and I stated that the term "chad" gained notice in popular culture because it was used in news accounts of the 2000 Election problem in Florida. I suspect that news journalists were more familiar with the term chad, because of the TTY equipment used by wire services. But I never heard an IBMer (other than myself) use the word "chad".

My father, now 94, is still living. I'll ask him how far back "chad" goes, in his memory.

--RussHolsclaw 15:27, 7 March 2006 (UTC)


 * I was concerned the wording gave the impression that chad was only used in connection with punch cards after 2000, which is not the case. The word chip would be a problem today, of course, because it's used for ICs, so that could be a reason the press settled on chad. This article should reflect both terms and there should be a mention of chip in the chad article. I suspected that chad was much older than 1947. You should also cite your teletype manual at that article. I would strongly urge you to get your dad to put his recolections about his work on tape. So much of this information is gettting lost. --agr 17:05, 7 March 2006 (UTC)

I talked to my dad on the phone tonight. He started working on Teletype gear in 1939, and the word chad was in common use at that time, by his recollection. As to the ambiguity between card chips and semiconductor ICs, I've often suspected that the little silicon rectangles may have been dubbed "chips" because of their resemblance to card chips, both in size and shape. Of course, I know of no way to confirm this, but it seems logical. They were also called "dice" (or, in singular form, "die") because the silicon wafer is "diced up" to make individual ICs. --RussHolsclaw 05:29, 8 March 2006 (UTC)


 * Usage at 3M. In the 1980s I maintained a line of engineering cameras, readers, printers and card duplicators (the 9x8 series) for 3M Co; The training, engineering and maintenance staff universally used the term 'chad' for the rectangular punchings.

LorenzoB 17:05, 6 October 2007 (UTC)

Univac also uses the term chips and the box that collects the chips is called the "Chip Receiver". Source: Univac 1700 Series Operating Instructions, published by Remington Rand Corporation, (c) 1969, 1970. --mikeu (talk) 23:09, 13 January 2008 (UTC)

I can assure younger folk that round punchings from paper tape were called "chad". I was the Flexowriter technician for the BMEWS DIP in Colorado Springs, before the NORAD COC went under Cheyenne Mountain. The USAF had data buffers that punched chadless tape from intermittent, bursty incoming data feeds. Slower readers had sense pins on the end of a rather long (2 ft?) swinging arm that "walked" up right next to the punch head if there was no incoming data. Blank tape rolls were comparatively enormous. Ordinry ones were maybe 9 or 10 in. dia, but these were more like a yard/meter! Regards, Nikevich (talk) 19:13, 10 March 2011 (UTC)

IBM and the Holocaust
Why theres no mention about the punched cards that the nazis in collaboration with IBM used to classifie the communists, jews, gypsies and others. This was one of the first apllications of this technology to track the victims of holocaust. Read: IBM and the Holocaust —Preceding unsigned comment added by 90.164.52.89 (talk) 13:41, 27 March 2011 (UTC)

Still in use
Punchcards are still used for places like carparks etc.
 * and time cards. RickK 03:54, 28 Jan 2004 (UTC)

As much as I agree that punch cards probably altered the 2000 election, that hasn't actually been verified, has it? Is that statement not a bit biased? (I'm referring to the "Hanging Chad" section.
 * Evanbro 00:43, Sep 8, 2004 (UTC)

They're also still in use to program fountains. Tragic romance (talk) 18:56, 31 October 2011 (UTC)

Fountain programs
The metal plates used to "program" water fountains, look like large punch cards. The ones I've seen are round, and somewhat smaller than a manhole cover. If looms are included, why not fountains? Tragic romance (talk) 19:07, 31 October 2011 (UTC)


 * Interesting. Worth mentioning, IMO, even if only as a minor point buried within the article. It also makes my mind turn (no pun) to the cam drums on music boxes. A metal disc storing the program for a fountain seems directly analogous, if the "reader" is a cam follower that opens and closes water valves directly. (Is it?) This is an interesting train of thought because it makes one ponder the essential similarity between punch cards and music box cams. The one differentiation that one can make is that punch cards fed data to a processor that could then do math upon it, whereas solely mechanical things like music boxes did not do any processing. But both media store data as holes in a sheet of material. Interesting. — ¾-10 01:43, 2 November 2011 (UTC)

age of card
In the section "IBM 80 column..." there is a photo of a card with the caption: "A general-purpose punched card from late-twentieth century." It looks to me that it is more likely to be much older, even late nineteenth century. Bubba73 You talkin' to me? 23:59, 25 November 2011 (UTC)


 * That section is about the IBM card designed in 1928. While it may indeed be old - as it appears to be, it is no older than 1928! 69.106.232.2 (talk) 10:24, 15 December 2011 (UTC)
 * And the card in question has rounded corners, a relatively late innovation (~1960s, IIRC).--agr (talk) 16:25, 15 December 2011 (UTC) Here is a photo from the 1950s showing square cut punched cards of the era: File:Keypunching at Texas A&M2.jpg.--agr (talk) 21:36, 15 December 2011 (UTC) And a ref from our article that is definitive: http://www-03.ibm.com/ibm/history/exhibits/supplies/supplies_5404PH13.html.--agr (talk) 17:07, 10 January 2012 (UTC)

Mohawk Data Sciences
Re: Mohawk Data Sciences introduced the first magnetic tape encoder in 1965, a system marketed as a keypunch replacement which retained the 80-character record and was immediately successful.. It wasn't the first. See UNITYPER. (Nor was the original machine very successful, although the Computer History Museum says it was.) --John Nagle (talk) 07:28, 28 January 2012 (UTC)
 * Good catch. I did not know about the Unityper. It seems Mohawk Data Sciences improved on that somehow (they got a patent for theirs) but I haven't seen a source that says how they improved on Unityper. So I added mention of Unityper and toned down the statement about MDS until a reference can be found that says more precisely what their specific innovation was.  Wbm1058 (talk) 18:08, 23 February 2012 (UTC)

500lb Keypunch?
The keypunch machine was a complex electromechanical device and weighed about 500lb, from article. I can't recall ever moving one, but this is hard to believe. Any source for the this? Else it should be removed. 69.106.254.246 06:12, 27 September 2006 (UTC)


 * I never moved one either, but those were big machines, and made almost entirely of steel. Steel frame, desk, steel covers over everything, and heavy-duty, industrial-strength motors, solenoids punch mechansim, printing mechanism, transport, etc.  I used to ride a motorcycle that weighed about 500 lbs.  I can totally believe that an 029 keypunch would weigh about the same. 206.210.75.84 (talk) 13:01, 4 June 2012 (UTC)

"Do not fold, spindle or mutilate" a "motto"
I saw the word "motto" in this sentence and thought to change it to "meme". But I checked out the citation for the sentence and it makes no mention of the phrase's cultural aspect and especially doesn't refer to it as a motto. What do you guys think? Is this something that needs to be more specifically cited or is it common knowledge? Makes me ponder the lifetime of common knowledge. How long until the common knowledge of the 80s and 90s is no longer common knowledge? — Preceding unsigned comment added by 71.238.28.104 (talk) 05:28, 13 October 2012 (UTC)


 * It certainly is common knowledge among those who were adults in the 1960s and 70s. "Motto" doesn't seem to be the right name for the expression, and "Meme" doesn't work, because the word didn't exist then. I'm desperately trying to come up with the right word. Might take it to the Language reference desk. HiLo48 (talk) 05:47, 13 October 2012 (UTC)


 * I did just that, and came up with catchphrase to replace motto. What do you think? HiLo48 (talk) 21:51, 13 October 2012 (UTC)


 * Sounds good to me. Just wanted to add, though, in reference to "... and 'meme' doesn't work, because the word didn't exist then": One certainly could identify memes from eras before that name for them was coined, just as one could talk about bacteria, or trees, that existed before humans coined the names "bacteria" or "trees". However, I agree that "catchphrase" works just fine. — ¾-10 16:16, 14 October 2012 (UTC)

Columns 73-80 were reserved ...
Looking at the IBM 704 manual, http://www.bitsavers.org/pdf/ibm/704/24-6661-2_704_Manual_1955.pdf, it states on page 39 that "Only 72 columns of the standard IBM card, however, can be read.". It's not a surprise then that the 1956 Fortran manual for the IBM 704 on page 8 states "Columns 73-80 are not read by FORTRAN and may be punched with any desired identifying information.".

How or where, with the passage of time, did that change to "Reserved for"? Where is the source for such a blanket statement? Computer installions may have made such a rule, but that rule would be for that installation, not for all punched cards.69.106.232.37 21:16, 29 September 2006 (UTC)


 * I changed the text, dropping "reserved for". Is that better? Use of sequence numbers was a wide spread practice . It's mentioned in Organick's Fortran IV Primer p.182. --agr 23:35, 30 September 2006 (UTC)

Yes, thanks. For me there is a big difference between "reserved for" and "could be used". Thanks again, 69.106.232.37 06:14, 1 October 2006 (UTC)


 * Yes, that was an IBM 704 thing, and it made it into FORTRAN. The 704 was a 36-bit machine, and the card reader transmitted the bits from each card to the CPU as two 36-bit words.  The 704 then had to "turn the corner" (convert the row-wise information to column-wise information by transposing the bit matrix) in software.  The hardware could read any 72 columns, if you wired the card reader's plugboard appropriately, but the usual setup was to read columns 1-72.  The 704 was a real stored-program computer, but the card peripherals and the printer were all based on plugboard-wired tabulating machines. IBM was still inching their way into all-electronic computing at that point. The 709 had the same set of clunky peripherals, but the 1401 and 7090, the transistorized machines, used completely different peripherals not based on old unit record equipment, and plugboards disappeared at last.  --John Nagle (talk) 06:04, 21 September 2009 (UTC)

In the 1980s, I encountered some old Fortran program decks wherein cards containing program statements had a "C" or an asterisk punched in column 80. Of course, the program ignored everything after Column 72. The idea was that this statement could be disabled by flipping the card over the long way. This places the comment indicator (the "C" or "*") in the first column, where it became effective, causing the program to ignore this card for processing purposes. The statement could be re-enabled by flipping the card back, and so on, and thus the program could be run in different modes without re-encoding anything. As the cards were symmetrical, they could be read either way. Of course, when a deck with reversed cards got duplicated, the resultant deck now had all the cards apparently in the same orientation, but some of the "comments" had peculiar, reversed text, like "C [many spaces] 2C ,1C ,B ,A TNIRP      ". WHPratt (talk) 17:46, 24 October 2014 (UTC)


 * As noted, the use of columns 1 to 72 for Fortran source, leaving 73-80 for other uses, originated on 36 bit machines, and was then included in the Fortran standard. More modern Fortran standards allow for free form (with different restrictions), but still allow for fixed form with columns 1-72, and continuation character in column 6. IBM then used a related convention for S/360 assembler.  (The actual columns can be changed with an ICTL assembler instruction, but that is more likely to confuse readers.  In the usual system, column 72 is where a continuation character goes, and the statement continues on the next card in column 16.)  S/360 utility programs also follow this convention. I believe COBOL also follows this convention, though I have never written any programs in it, and don't know the history of it all that well.

Hollerith encoding
- Hollerith encoding was a standard for punch cards used in early computers. Until the early 1980's, the use of these punch cards was a common form of computer input. - These punch cards were 80 columns by 12 rows in size, and in the Hollerith encoding scheme, had one BCD character per column.  - An artifact of this early "standard" is that most character-based terminals used an 80-column by 24-row size. Even now, the default size for character interfaces remains set at 80 columns.

I merged the above here from Hollerith encoding, intending to integrate it into the article, removing redundancies, but ran out of time. Yes, it's largely redundandant, that's why it's being merged. But something should be said. --SmokeyJoe (talk) 22:43, 28 March 2009 (UTC)


 * Is it so obvious that 80 column terminals derive from 80 column cards? At 10cpi, a printer prints 80 columns in 8 (US) inches, convenient fit for 8.5in paper with a small margin. Or were cards designed as 80 columns to fit existing printers? Gah4 (talk) 19:39, 6 May 2015 (UTC)

External links modified
Hello fellow Wikipedians,

I have just added archive links to 1 one external link on Punched card. Please take a moment to review my edit. If necessary, add after the link to keep me from modifying it. Alternatively, you can add to keep me off the page altogether. I made the following changes:
 * Added archive https://web.archive.org/20150812004449/http://www.washingtonpavilion.org/VisualArtsCenter/events/punchcardart.cfm to http://www.washingtonpavilion.org/VisualArtsCenter/events/punchcardart.cfm

When you have finished reviewing my changes, please set the checked parameter below to true to let others know.

Cheers. —cyberbot II  Talk to my owner :Online 10:50, 27 August 2015 (UTC)
 * Fixed archive by linking to an earlier version. Mojoworker (talk) 21:11, 29 August 2015 (UTC)

Use in toys
Weren't punched cards or tape used in some toys slightly before modern microelectronics? -- Beland (talk) 20:15, 26 September 2015 (UTC)

Other languages
Why can't I see links to another languages from English page? — Preceding unsigned comment added by 95.25.205.247 (talk) 00:06, 1 November 2015 (UTC)

Why is the punched card the size of the 1887 US dollar bill? - part 2.
See Talk:Punched_card

The following has been removed from the article - the main page is not the place for a discussion.


 * 1) Did Hollerith envisage using currency handling eqipment and boxes for punched cards?  Citations please.
 * 2) Smaller decks of cards were not kept in the large boxes (my personal experience) - citations for this also please.

Can we sort this out here, then update the main page accordingly. Martin of Sheffield (talk) 10:02, 16 March 2017 (UTC)

(per discussion on alt.folklore.computers, cards are not ascii, ebcdic, etc.)
Someone noted: (per discussion on alt.folklore.computers, cards are not ascii, ebcdic, etc.)  If a card contains only printable characters and blank space, I would 100% agree. EBCDIC defines 256 punch codes, such that rows 1 through 7 have zero or one punch, which conveniently leaves 256 combinations. There are some printable characters, and many control characters, that don't have a correspondence between ASCII and EBCDIC, and even less, any other punched card code. Seems to me that cards using the 256 codes, such as OS/360 and successor object programs, can reasonably said to be using EBCDIC. At the time when S/360, and EBCDIC, were being developed, there was a proposal for an ASCII-8 code. That is, an eight bit extension to ASCII that isn't just ASCII-7 in the first 128 positions. I believe that all 256 have defined punch codes. But the standard never was approved, and as far as I know, hardware never developed to use it. (Other than that S/360 machines can generate appropriate signs for decimal arithmetic and zones for UNPK. But that is only 10 of the 256 code points.)  Gah4 (talk) 19:30, 21 March 2017 (UTC)


 * There is a direct mapping between later punched card codes and EBCDIC. See . John Nagle (talk) 19:47, 21 March 2017 (UTC)


 * But are there other codes with 256 code points that map to 256 punch code points? I know that DEC sold card readers and card punches for some of their computers.  Is there a mapping between the VAX/VMS character set and card punch code points?  Gah4 (talk) 20:02, 21 March 2017 (UTC)

Cultural impact
When I was growing up in the mid-70s punchcards were already on their way out. Still, I remember how used cards were incorporated into arts and crafts projects--I remember making Christmas wreaths in cub scouts and somewhere there were folks making lampshades. My father coached high school football and I remember that the confetti people threw at the games had tiny numbers printed on it--they were the bits that had been punched out to make the holes in the punch card. I suppose they were collected and made some companies a little extra money. PurpleChez (talk) 19:53, 13 March 2011 (UTC)

I used to have a ballot for the 1969 Major League Baseball All-Star Game, which I didn't submit but retained as a bookmark. http://www.uni-watch.com/wp-content/uploads/KGrHqRnQFKnh8KfBQVfidM1Q60_57.jpg This was the year that the vote was returned to "the fans" (i.e., the general public) after a decade or more of the vote being restricted to the players. You could pick these up at drug stores and such, so ballot-box stuffing was still possible, and indeed practised, though on a smaller scale than before. It was a standard 80-column Hollerith card, with text printed on it to identify the marks. The voter was expected to punch out holes next to the names of his/her choices. Of course, the tabulating machine was programmed to identify the candidate via the row and column positioning of the punch. Years later I ran the card through a keypunch in the computer lab just to prove that it would fit. This might be a good example of the popular use of such cards. WHPratt (talk) 12:45, 23 June 2017 (UTC) Edit: Okay, the picture does show a 1973 ballot, but I'll bet that the 1969 card was of similar structure. WHPratt (talk) 12:49, 23 June 2017 (UTC)
 * It may just be my, but all I'm seeing is "403 Forbidden". Martin of Sheffield (talk) 13:00, 23 June 2017 (UTC)
 * That's what I got at home, though it worked at the office! Sorry about that. Just do a Google on "1969 Major League All Star Ballot" and select "Images." You'll see several such ballots. WHPratt (talk) 00:25, 24 June 2017 (UTC)

References for a 2017.6.30 edit to punched card. IBM name not on products until 1933
https://books.google.com/books?id=OyfOAAAAMAAJ&q=80+%22punched+card%22&dq=80+%22punched+card%22&hl=en&sa=X&ved=0ahUKEwj58oPX7-bUAhVF1WMKHSk6CO0Q6AEIKzAB

Proceedings - American Gas Association, Volume 11 Amaerican Gas Association, incorporated, 1929

p.223 Each machine system possesses the advantage of certain mechanical refinements, but the essential difference between the two punched card billing systems is that under the Hollerith method the tabulating cards are first punched with all the billing data and proved ...

... or have definityly decided to use the Hollerith punched card billing system. The smallest company issues approximately 22,000 bills per month -- the largest 475,000. Three companies are using 80-column equipment; the others are using 45-column equipment. In addition, on other utility is experimenting with 80- ...

-- https://books.google.com/books?id=O4AYAQAAIAAJ&q=80+hollerith+%22punched+card%22&dq=80+hollerith+%22punched+card%22&hl=en&sa=X&ved=0ahUKEwjB8-7l_ObUAhXn5oMKHYkVA58Q6AEIKzAB

The Record, Volume 20 The Institute, 1931

p.337 In his system of maintaining the records and accounts, the author makes use of the addressograph, Hollerith 80 column punched card ...

p.378 Of interest to companies using the Hollerith 80 column equipment ... 67.160.196.6 (talk) 02:33, 1 July 2017 (UTC)

Port-A-Punch details
Is it worth mentioning that the Port-A-Punch only permitted 40 columns to be punched (skipping alternate columns to keep the card somewhat stronger), or that it was used for coding and storing programs for the Monroe 1665 Programmable Calculator (presumably under license)?: http://www.oldcalculatormuseum.com/monroe1665.html, http://www.oldcalculatormuseum.com/monpcard.jpg —Preceding unsigned comment added by 75.17.114.137 (talk) 08:12, 10 December 2009 (UTC)


 * You'll be pleased to know - 40 column text added 16:58, 3 April 2017‎. 67.160.196.6 (talk) 19:22, 3 July 2017 (UTC)

Inquiry on corner-cuts
I worked for IBM for a number of years as a Customer Engineer, trained on the inner workings of a number of punched-card machines made by IBM. During that time, I never saw a machine that had a "corner cut switch" to stop the machine when a corner cut was either present or absent. The machines just didn't have any corner-cut detection hardware at all. The only exception to this I recall was that some machines, notably card sorters, had a special extra-cost feature called "master/detail", which could use corner cuts to distinguish a "master" card (with information about, say, an account), and "detail" cards that were inserted into the deck behind their respective "masters". When turned on, the effect of this master/detail detection hardware was to cause the detail cards to be automatically sorted into the same pocket as the last master card detected.

Does anyone out there have any specific information about the allegation that machines had a corner-cut detector for the purpose of detecting cards that were not oriented correctly? If so, please provide details, such as the make and model (or IBM machine-type number) of the machine in question. It seems to me that this assertion in the article may have been an uninformed conjecture on the part of the person who entered it, and is not backed by any true knowledge. --RussHolsclaw 18:05, 5 March 2006 (UTC)


 * Yes. Special feature 4714 for IBM model 24 and 26 keypunches supported interspersed gangpunching based on detection of upper left or upper right corner cuts.]. This was presumably for shops that didn't have a reproducer like an IBM 514, the usual machine for gangpunching. --John Nagle (talk) 05:37, 21 September 2009 (UTC)


 * The exact answer to your question is No, there was no feature to detect incorrect orientation.


 * But there were machines that could recognize - and act - based on corner cuts. In addition to the keypunch (above) an optional feature on some sorters (see A24-1010 Operator's Guide p.42 for example) was a "Card Matching Device" that could sort master cards from detail cards - determined by the card's corner cut (or by a punch in col. 1 or 80).  67.160.196.6 (talk) 19:02, 3 July 2017 (UTC)


 * The usual feature for ensuring cards were orientated correctly in the reader was a mk1 eyeball and thumb. Martin of Sheffield (talk) 19:31, 3 July 2017 (UTC)

Inspiration for Hollerith card
What I was told was Hollerith's inspiration to use punched cards for the census data was observing a train conductor punching tickets (a punch in one spot for a man, another for a woman, another for a child, another for something else, etc.) so that when he came back again and checked tickets another time he could inspect the punches and verify that the ticket was for the right passenger, destination, etc. and had not been "passed on" to someone that hadn't bought it. Hollerith decided he could use a similar coding system for census data, but would need some level of mechanical assistance to punch and read the many holes he would need (the conductor only needed to punch a small number which could be verified at a glance). -- RTC 01:58, 16 November 2006 (UTC)


 * "train conductor punching ..." is correct. See Austrian's book p.15 for Hollerith's text.  The complete Hollerith letter is (or used to be) accessible online. 67.160.196.6 (talk) 23:01, 3 July 2017 (UTC)

IBM Votomatic
The article text asserts that the Votomatic card was a Port-A-Punch card. Is that know to be true, or is it an assumption based on similar construction of the cards? If true, then IBM Votomatic is an application and should not have its own heading under Card Formats. 69.106.232.37 15:22, 10 October 2006 (UTC)

True. Votomatic is now listed with card handling machines. 67.160.196.6 (talk) 00:54, 4 July 2017 (UTC)

References for a 2017.6.30 edit to punched card. IBM name not on products until 1933
https://books.google.com/books?id=OyfOAAAAMAAJ&q=80+%22punched+card%22&dq=80+%22punched+card%22&hl=en&sa=X&ved=0ahUKEwj58oPX7-bUAhVF1WMKHSk6CO0Q6AEIKzAB

Proceedings - American Gas Association, Volume 11 Amaerican Gas Association, incorporated, 1929

p.223 Each machine system possesses the advantage of certain mechanical refinements, but the essential difference between the two punched card billing systems is that under the Hollerith method the tabulating cards are first punched with all the billing data and proved ...

... or have definityly decided to use the Hollerith punched card billing system. The smallest company issues approximately 22,000 bills per month -- the largest 475,000. Three companies are using 80-column equipment; the others are using 45-column equipment. In addition, on other utility is experimenting with 80- ...

https://books.google.com/books?id=O4AYAQAAIAAJ&q=80+hollerith+%22punched+card%22&dq=80+hollerith+%22punched+card%22&hl=en&sa=X&ved=0ahUKEwjB8-7l_ObUAhXn5oMKHYkVA58Q6AEIKzAB

The Record, Volume 20 The Institute, 1931

p.337 In his system of maintaining the records and accounts, the author makes use of the addressograph, Hollerith 80 column punched card ...

p.378 Of interest to companies using the Hollerith 80 column equipment ...

67.160.196.6 (talk) 02:33, 1 July 2017 (UTC) — Preceding unsigned comment added by 67.160.196.6 (talk)


 * None of the above any relevance to adding "A mobile Hollerith (BTM) installation was landed on the Normandy beaches on D-Day plus six.". It's a non-notable event, and has no bearing any path of development of punched cards; it's just trivia.  It's placement simply disrupts the prose of the section.  Also, please follow WP:BRD (note it's not "BRRD"); reverting a revert at this point is WP:EW.  Wait until discussion reaches consensus to make changes.  The proper procedure at this point is to undo the 2nd (disruptive) revert and restore the the section to it's pre-contention state at this time.  --A D Monroe III (talk) 13:41, 17 September 2017 (UTC)


 * The text included above pertained to the correcting of "In 1931 IBM began introducing multiple punches for upper-case letters and special characters" NOT to the D-Day edit.67.160.196.6 (talk) 07:20, 6 December 2017 (UTC)


 * As for D-Day 1) the history is chronological and 2) it is an example of Punched card machines where you would have never expected to find them - thus punched cards are everywhere.67.160.196.6 (talk) 07:20, 6 December 2017 (UTC)

And while you're here can you help with "In 1881 Jules Carpentier ..." please?
The article has a beautiful first line"A punched card or punch card is a piece of stiff paper that can be used to contain digital information represented by the presence or absence of holes in predefined positions"

Unfortunately someone added text to the history section begining "In 1881 Jules Carpentier developed a method of recording and playing back performances on a harmonium using punched cards." Music recorded with punches does not meet the "predefined positions" criteria. Not meeting that criteria means the recording is not digital. I deleted, but was reverted.67.160.196.6 (talk) 07:20, 6 December 2017 (UTC)


 * The currently popular player pianos with music on rolls of paper use analog timing tracks. I don't know the details on the Carpentier system, so it might be analog or digital. I don't know that analog should exclude it from this article, though. Gah4 (talk) 17:49, 12 June 2018 (UTC)

Pre-decimal
What does a punch card have to do with a pound sterling. This seems quite specious. — Preceding unsigned comment added by 73.176.108.118 (talk) 03:59, 24 December 2017 (UTC)


 * It illustrates how an apparently decimal based system was adapted to handle other radices. BTW, please remember to sign your edits.  Regards, Martin of Sheffield (talk) 10:49, 24 December 2017 (UTC)


 * The early PL/I compilers have a data format, and conversion ability, for Pound Stirling data. I am not sure now what the external format is, though. I suspect, like many PL/I features, it came from COBOL. Gah4 (talk) 17:42, 12 June 2018 (UTC)


 * According to IBM Operating System/360 PL/I: Language Specifications, the form used in PL/I programs was {pounds}.{shillings}.{pence}L, e.g. "1.10.0L". The spec says "Sterling fields are considered to be real fixed-point decimal fields. When involved in arithmetic operations, they will be converted to a value representing fixed-point pence.", so internally, and presumably when written out in binary form, they may have been packed decimal numbers of pence with conversion done as part of I/O, or N digits of decimal pounds, 2 digits of shillings, and 2 digits of pence, with conversion done to and from pence in arithmetic (my bet's on the former).


 * What I forgot, though, is what the external representation looks like. It seems (page 149) that there is single character representation for shillings and pence using BSI specification, and single character representation for pence using IBM representation. Gah4 (talk) 16:08, 13 June 2018 (UTC)


 * By the way, this is an interesting manual, that defines the language independent of what is implemented by any compiler. That is, the way it should be, not necessarily the way it is. Gah4 (talk) 16:08, 13 June 2018 (UTC)


 * It's interesting that the sterling picture section includes "P Specifies that the associated field position contains the pence character P." In 1965 the symbol pence was "d", using "P" would immediately mark you out as ill-educated or foreign. Martin of Sheffield (talk) 16:44, 13 June 2018 (UTC)


 * I guess this was before ownership of PL/I went to Hursley, then. :-) Guy Harris (talk) 17:35, 13 June 2018 (UTC)


 * I couldn't find a COBOL 60 spec online; there might have been support for sterling in the spec, or it might just have been an enhancement added by UK vendors. Guy Harris (talk) 07:26, 13 June 2018 (UTC)


 * £sd doesn't seem to discuss the one character representation, so I asked in talk. Gah4 (talk) 16:17, 13 June 2018 (UTC)


 * I found Collected Information on Punched Card Codes which seems to be from 1960, with some discussion of BSI codes. First, they number the top zone 10, instead of the IBM 12. That allows for rows 0 to 11, which they use for pence and months, with 0 for December. I believe, then, with a 10 zone, and 0 though 9 digit punch, for 10 to 19 shillings. Gah4 (talk) 17:02, 13 June 2018 (UTC)

Control vs. data, continued
The other one was getting long, and I now actually removed Prior uses of machine readable media, such as those above (other than Korsakov) had been for control, not data. It does seem that control vs. data is important, but I think this isn't correct. Gah4 (talk) 19:03, 15 June 2018 (UTC)

Collected Information on Punched Card Codes
While looking for something else, I found: Collected Information on Punched Card Codes. It seems like a useful reference, and maybe some should go into the article. Gah4 (talk) 16:28, 13 June 2018 (UTC)
 * Nice find. I've added it to the external links section.--agr (talk) 20:52, 15 June 2018 (UTC)

control vs. data
The statement Prior uses of machine readable media, such as those above (other than Korsakov) had been for control, not data. doesn't seem quite right. Loom patterns are data. On the other hand, loom patterns are not for data processing, but data storage. Gah4 (talk) 17:39, 12 June 2018 (UTC)


 * "Loom patterns are data." The History section begins by saying "Basile Bouchon developed the control of a loom by punched holes in paper tape in 1725." - emphasis mine.  That was an early example of data controlling a machine (the program into which I'm typing this is a later example...).


 * I'd say the only stuff that's purely data is stuff that's processed but not "interpreted" in the way machine code, microcode, bytecode, etc. is. Guy Harris (talk) 05:16, 13 June 2018 (UTC)


 * If I print a file of numbers on a mechanical printer, the numbers control mechanical positions of type wheels or bars or pins, but it is still data. Control characters, like newline, might not be considered data. As well as I know, the punched cards for the loom are only data (what color to put where), and not control (there is no loop structure, or GOTO). Maybe there is a stop code, though. Gah4 (talk) 16:46, 13 June 2018 (UTC)


 * It all depends upon your POV. To a programmer GOTO is a control statement, part of the language.  To a compiler GOTO is part of its data stream which it processes and outputs as another form of data (in this case machine code). Martin of Sheffield (talk) 16:57, 13 June 2018 (UTC)


 * Yes. But also, numerical data in compiler input often comes out slightly processed in the output. String data likely comes out exactly the same. Numeric data converted from ASCII (or EBCDIC) to binary integer of floating point values, but still data. Gah4 (talk) 17:09, 13 June 2018 (UTC)


 * Is the ability to iterate other than by organizing the instructions in a physical loop, or to make decisions, a requirement for some form of data to be considered "control"? The Harvard Mark I page claims that "At first, the Mark I had no conditional branch instruction. This meant that complex programs had to be physically lengthy. A program loop was accomplished by joining the end of the paper tape containing the program back to the beginning of the tape (literally creating a loop)."


 * And, at least as I read the Jacquard loom page, the punched cards don't directly encode colors, they encode whether the warp thread is lifted or not, to put the weft thread above or below it.


 * I'd consider a catalog entry that just says "this is an ugly Christmas sweater with a drunken Santa Claus" to be data and not control, but something that can instruct a machine to produce that sweater to be data used as control. Guy Harris (talk) 17:51, 13 June 2018 (UTC)

I think the distinction you guys are trying to make is not so simple. The data in a loom card selects the color of a particular weft thread and which portion of it is to be visible. (Ada Lovelace mentions that "There is in existence a beautiful woven portrait of Jacquard, in the fabrication of which 24,000 cards were required.") The data is translated into the action by the holes, which control which warp threads are lifted. So there is both a data and a control aspect. But the story is the same for Hollerith punched cards used with early electro-mechanical accounting machines. The hole pattern in Hollerith code is not accidental, it was developed to allow electro-mechanical machines to accumulate totals. In a field that is to be added to an accumulator (as determined by control panel wiring) the presence of a numeric punch allows a counter wheel to start rotating, one step per card row, and the counter stops when the card being read advances to the zero row. Thus a 9-punch allows nine counter steps to accumulate, an 8-punch only eight, and so on. That is why cards were always fed 9-edge first. So just as with the loom, there was both a data aspect and a control aspect to Hollerith punched cards. Of course with later electronic systems, the card coding was separated from the control action, but for most of punched card history that was not the case.--agr (talk) 19:19, 13 June 2018 (UTC)


 * The hole patterns for Hollerith don't make it not data, but a convenient data encoding. The tied together cards in a Jacquard loom don't allow for a branch operation, which would be control. Well, I could ask if there is a WP:RS about Hollerith census cards being data and not control. Otherwise, is there an important distinction between loom pattern data and census data that needs to be considered? Gah4 (talk) 20:47, 13 June 2018 (UTC)


 * "The tied together cards in a Jacquard loom don't allow for a branch operation, which would be control." I am not yet convinced that branches (other than looping, where a single loop can be done simply by literally making a loop...) is a prerequisite for something being deemed "control".


 * "Otherwise, is there an important distinction between loom pattern data and census data that needs to be considered?" Whilst the census punched cards did, at a very low level, control pulses that operated counters, they didn't specify which counters to activate, or any other aspect of the process of counting the census results; the loom punched cards did specify the process of weaving the cloth.  In the machine on which I'm typing this, the bits in a number from a GPR or a register into which memory data is fetched will, when gated to an ALU, turn on or off transistors in the ALU depending on the bit values, but I don't deem that to be "control" at the level of instructions fetched from memory.  I'm as yet unconvinced that the low-level fashion in which punched cards control counters should be considered "control", or that if it's not considered "control", the fashion in which punched cards control a Jacquard loom should not be considered "control", either, so you might need a reliable source for that claim.  Guy Harris (talk) 23:00, 13 June 2018 (UTC)


 * I wasn't suggesting that punched card are not data, quite the contrary, I was arguing that loom cards, even though they have a control function are data every bit as much (pun intended). The 24,000 loom cards that produced the image of Jacquard that Lovelace mentions constitute nothing less than a digital image. The pixels are exposed wefts crossing individual woof threads. If one were able to read the cards into a computer, with 1 and 0 bits representing the present or absence of a hole in each position, it would be straightforward to convert that data into, say, a TIFF file, if one had a basic understanding of the loom set up and the colors of the threads used. Even without knowing anything at all about the loom, there would be enough information in those 24,000 cards to recover the image using cryptanalytic techniques. That set of loom card is certainly a dataset.--agr (talk) 15:31, 15 June 2018 (UTC)


 * Yes, I believe that loom cards are mostly data. Conditional branches are a common type of control function, that I don't believe was used by Jacquard. As above, I suspect that a STOP signal would be convenient, and that would be control and not data, but just one bit of control. Gah4 (talk) 18:57, 15 June 2018 (UTC)
 * Looking at YouTube videos of Jacquard looms in operation, it seems the operator propelled the shuttle back and forth and would have had the ability to stop at any point, say if a thread broke. Looping mechanisms to control weaving were know before Jacquard and were used for simpler patterns that repeated. But punched cards did not have conditional branching either, not counting their much later use as a medium to enter computer programs, which just used the wide availability of relatively inexpensive keypunch machines to serve as a text entry medium . There were limited conditional abilities, say to to print the totals of the previous detail cards when a new master card is detected, but no real branching. Incidentally the Jacquard image I mentioned is available in commons as File:A la mémoire de J.M. Jacquard.jpg and is incorporated in the Jacquard loom article.--agr (talk) 03:33, 17 June 2018 (UTC)


 * "There were limited conditional abilities, say to to print the totals of the previous detail cards when a new master card is detected" That sounds as if you're talking about accounting machines; I'd consider their control to be done solely through the plugboard, with the punched cards being purely data. Guy Harris (talk) 08:25, 17 June 2018 (UTC)
 * Yes, I am talking about accounting machines. They were the primary way punched cards were processed for most of the punched card era. By the time we get to computers (with some early exceptions) cards were just data, even cards containing program source and object code. But while accounting machines were indeed controlled by plugboard wiring, that wiring could and often did have options that were selected by special coding on the cards. There was more than one read station (at least on the later accounting machines), so that, for example, the accounting machine could detect a punch indicating a master card and then print totals from the previous detail cards before processing the new master card. One might also have a special control card to be placed at the end of a deck whose only purpose was to trigger printing grand totals for that run. People using these machine were only trying to get their work done. There were many situations and a lot of ingenuity was employed to get the machines to do what was needed. I don't think our article should emphasize a distinction that was not so clear cut at the time.--agr (talk) 15:39, 18 June 2018 (UTC)

2003
Why is it "punched tape" and "punch card"?
 * It is punched card Pi314m (talk) 19:35, 5 November 2018 (UTC)

Hollerith's key punch codes system (zones and values) was implimented as FIPS-14. Just thought it would be nice to mention it here. I have never been able to get FIPS-14 our of my mind, even though I haven't used it in over a decade. Thanks, Matthew Brown Lake Oswego, OR

some material here moved from "Hollerith" and "Hollerith card", now redirects. See those pages for history. -- Someone else 05:46, 4 Nov 2003 (UTC)

Termnal screens
The article suggests that the 80 column card led to the use of 80 column wide terminals. It seems to me about as likely that 80 columns fit nicely at 0.1in spacing across the US letter format paper. Or maybe that 80 column wide paper led to 80 column punched cards? Gah4 (talk) 19:02, 9 September 2015 (UTC)


 * No comments on this one. I suspect that 80 column (after margins) typewriter pages came before 80 column punched cards, and later lead to 80 column computer printers and terminals. It is a convenient width for human readers, not a descendant of 80 column cards. Gah4 (talk) 20:06, 15 June 2017 (UTC)


 * It is much more likely that terminals took their size from punched cards than from a paper size that was not used in the computer industry. Standard paper width in the computer industry during mainframe and minicomputer era was 132 column sprocket fed.  The alternative to punched cards were initially "teletype-type" typewriters which also used 132-column output.  You wrote in 80 column, then compilers and linkers would add source line and variable information in the remaining 52 columns when you got your job back from the system.  As VTs started to become more common they were viewed as "glass teletypes", remember this was at a type when even VTs used line editing, not whole screen.  Later VTs like the DEC VT100 and its descendants gave you an ability to switch between 80 or 132 to match card and paper sizes.  Indeed even when lasers started to come in during the late '80s they were often configured to print in landscape mode so that 132 column output was rendered properly.  I suspect it was the rise of the cheap 80-column dot-matrix printers and the pricier daisy wheel printers coupled to PCs that really shifted the industry and brought secretarial attitudes to DP. Martin of Sheffield (talk) 20:53, 15 June 2017 (UTC)


 * Printing terminals trace back to predecessors of Teletype earlier than 80 column punched cards. The pictures look like 80 column width. The Teletype article doesn't indicate 132 columns until the sprocket feed model 43 in 1977. So, glass teletypes would have inherited the 80 column width though the whole line of Teletypes and predecessors, and card width through a similar lineage. Gah4 (talk) 22:13, 15 June 2017 (UTC)


 * The IBM golfballs I used in 1976 were 132 column sprocket fed, as were the DEC LA36s introduced in 1974. Interesting though the earlier DEC LA30 was only 80 column.  Smaller VAX and PDP installations often used LA120s as system printers if the cost of an LP25 could not be justified.  The LA36s were a right pain to use, at 30 character per second you could easily type faster than the computer, whereas an LA120 on a 9,600 baud serial line at least kept up with you. Martin of Sheffield (talk) 22:31, 15 June 2017 (UTC)


 * Yes, but 10 character/inch typewriters and 8.5inch wide paper trace back earlier than the 1928 punched card. It seems to me more likely that later terminals trace to those than to the card width, and that the 80 column card, also traces back through those. In other words, 80 columns is convenient for human readers and writers in general, which lead to a variety of 80 column wide media. Interestingly, from letter paper this history of the 8.5 inch seems to have been lost, but likely traces back to hand made paper. It seems close to the optimal width for human vision at a comfortable reading distance. Also, 80 columns is about the resolution for higher quality CRT displays from the beginning of the glass terminal era. Even higher resolution might have been available at much higher cost. Larger CRTs could give a larger display, but about the same pixel width. Broadcast television has much lower resolution, giving the 40 character width of early Apple displays, when used through a video modulator. The Processor Technology SOL and VDM-1 use a 64 character wide screen, possibly as it makes addressing easier. Gah4 (talk) 23:18, 15 June 2017 (UTC)
 * You're quite right about the TVs. I had a ZX80 and later a Spectrum and they used 32 columns with the characters displayed as 8x8 pixel cells.  I'll also accept that paper width may have influenced the card width (though I'd like to see evidence for that), but you need to remember that early terminals were used exclusively by DP staff and programmers, not by ordinary folk reading text.  80 is also way to convenient as a number!  A typist setting up her paper would typically use at least 1/2" margins which puts the line length at 75 characters.  80 is more the sort of number an engineer might land on.  I've just had a look at my old NAG mark 9 mini-manual from 1981.  It is printed using a typewriter font (as was typical of computer manuals at that date) with 1" margins.  My copy is on A4, but translating that to US paper sizes would indicate a 65 character line.  Using my printout ruler the lines are formatted with 5 characters for the paragraph number and the longest line is 59 characters giving 64 total.  A Calcomp manual from 1985 is also 5+59 but this time set on US letter paper (8.5").

There is a lot of speculation here, but it ignores the direct relation between the screen width of computer terminals and punched cards: a major early use for the early video terminals was in keypunch replacement via key-to-tape and, later, key to disk systems. By the 1960s, computers had largely replaced Unit record equipment and punch cards were mainly used for data entry in batch processing systems. The day's transactions were keypunched, verified and the cards were then read and written onto magnetic tape, often on a smaller computer such as an IBM 1401. The data on the tape was sorted on the computer (using several tape drives as intermediate storage) and then used to update the master file tape and produce reports. Key-to-tape systems, e.g. by Mohawk Data Sciences and Inforex, replaced a group of keypunches and verifiers with a set of computer terminals connected to a central computer that had an IBM-compatible 9 track tape drive. There was a big pent up demand for such systems, but early video terminals were much more expensive than keypunch machines, even including the ongoing cost of the cards. The breakthrough was integrated circuits, which allowed video terminals that were inexpensive enough to compete. A video terminal that displayed fewer than 80 characters would miss out on this major market, having more than 80 characters increased costs and offered little advantage at first. Wikipedia does not yet cover this era in computer history well. --agr (talk) 10:59, 16 June 2017 (UTC)


 * Yes, lots of speculation an no WP:RS that 80 column terminals followed directly from 80 column cards, instead of both having a common origin, or that 80 just makes sense. It is a little easier to build 64 character width, as it can be easily counted with a six bit counter, but it is just a little bit too small. I suspect that many Teletype machines are 72, which with margins is a reasonable value.  The choices for terminals might have been 72 or 80, and maybe card width was enough to tip toward 80. Or maybe the extra ability, and market competition led to 80, if the cost difference isn't all that much.  The IBM 2260, a video terminal from 1964, started out at 40 characters wide, with the model 3 at 80. Yes, within IBM 80 is a special number, but it isn't so obvious outside. Gah4 (talk) 20:33, 16 June 2017 (UTC)


 * If someone today talks about sending a message in 140 characters, everyone knows they are talking about Twitter. No one would suggest that they had some independent reason for picking 140. Twitter has only been in existence for 11 years. The 80-column punched card was in existence for over 40 years when computer terminals were being developed and totally dominated data processing for most of that time. The 80 character format was as closely identified with punched cards as 140 is with Twitter, even more so. The difference between a 64 counter and an 80 or 72 counter is one flip-flop and one and gate, a trivial cost even in the discrete transistor era. The high cost item for raster video terminals was sufficient memory to store all the characters on the screen. BTW, here is a source for the Datapoint 3300 having a 72 character line width to be compatible with Teletype terminals it intended to replace. The later Datapoint 2200 model went to 80. --agr (talk) 10:48, 22 June 2017 (UTC)


 * I have known that there was a reason behind the 140 character limit for Twitter, based on a limit for SMS, but didn't know more than that. It seems, from the SMS page that the limit traces back to GSM in 1984, at 128 bytes, or 160 seven bit ASCII characters.  (I don't know about 160 vs. 140.)  That traces back to the length of typical postcards and Telex messages. So, someone tried to understand a convenient limit for human messages. Card width and terminal width are different, as we are allowed to use multiple lines. There are widths that are more optimal for human vision, that are used for printing books. If you make the line too long, it is too hard to find the beginning of the next line, following the text back. That seems to be around 65 to 75.  The column width for newspapers is closer to 40, maybe it is easier to follow the flow with the smaller print typical for newspaper articles. Screen memory, and its cost, has interesting effects. The VDM-1 (popular S100 bus display device) and it successor, the SOL terminal, have a 16x64 display, using 1Kx1 SRAMs for display. If you increase to 24 lines, and 2K characters, you could have 85 character lines, and with 25 lines (24 plus a special message line) you get 81. It seems, then, that there are multiple reasons for 80 instead of 72. A 24x80 display also works well with video monitors available in the 1970s, for somewhat reasonable prices. Higher resolution cost a lot more. And as I noted before, 80 characters at 0.10in pitch fits well across on 8.5in wide page, with some margins. It seems to me, then, that it is the combination of all these reasons, that 80 characters became popular for later displays. Gah4 (talk) 08:31, 23 June 2017 (UTC)
 * So far, I don't see much discussion about where the 80 column card came from. It seems that they took the original size of the Hollerith card, and put more, narrower, columns onto it. Narrower columns make it harder to squeeze the read electronics (or optics) in that space, accurately enough, especially for the technology of the time. Gah4 (talk) 08:31, 23 June 2017 (UTC)

I found a source and added a paragraph on how the 80-column format came to be. Note that card readers at the time used wire "brushes" to read the cards. As the card passed under the brushes, the wires could fall into a hole and make contact with the metal surface below. A rectangular hole is better suited for this method of reading. As for the the source of the 140/160 character limit, our SMS article makes clear that it was a technical limitation: "However, it was necessary to limit the length of the messages to 128 bytes (later improved to 160 seven-bit characters) so that the messages could fit into the existing signalling formats. Based on his personal observations and on analysis of the typical lengths of postcard and Telex messages, Hillebrand argued that 160 characters was sufficient to express most messages succinctly."--agr (talk) 12:15, 30 June 2017 (UTC)


 * BTW, the reason for 80-column screens is based on software compatibility, not hardware. The first terminals were designed to exactly replace the functionality of a card reader, and used the same interpreter software; the format of a line on the terminal was identical to the text of a punch card, even including things like starting with "$JOB" and "$END" to distinguish command "cards" from data "cards" (each line on a on a terminal was even called a "card" for a while).  So, since cards were limited to 80-columns, the terminals duplicated this limitation.  Note that terminals were generally not 8.5 inches wide, so 80-char terminals had nothing to do with 10-chars-per-inch directly -- only indirectly via punch cards.  Unfortunately, I have no RS for this; I just experienced this transition first-hand, working on some very old computers when I was very young.  --A D Monroe III (talk) 18:12, 30 June 2017 (UTC)


 * You don't say which year that was, but the Baudot code, from which the term baud was derived, goes back to 1874. See teleprinter for the history of printing terminals in the late 1800's and early 1900's, and finally leading to the Teletype Corporation in 1928.  As far as I know, 8.5in paper and 10 characters/inch traces back pretty early, too.  Many Teletype machines print 72 columns on 8.5 inch paper, leaving good sized margins.  The IBM 2315  printer, used as a console terminal on many IBM computers, seems to be 126 characters wide.  Many IBM line printers print 132 columns (10cpi) across 14 inch (plus perforation) fan-fold paper. By the time that CRT terminals came along, there were a variety of printing terminals to emulate. The spot size of many monitors allowed for between about 64 and 80 characters across. Some early models were upper case only, with a 5x7 character cell. Later upper/lower case, 7x9 cell, plus descenders. (One popular character generator stores 7x9 (63 bits) plus one bit to lower the character two rows, for 64 bits/character. A convenient size, plus a small amount of extra logic.)  And, as I mentioned above, with a 24 line display and 2K character display memory, 80 is a convenient width. Many DEC terminals are 24x80, with a 25th status line, for 2000 characters.  With 2048 character memory, you could make a 24x85 screen, or 25x81. So, there were a variety of reasons for 80, and some for 72, and it might be that the 80 character card tipped the balance in the end. Each terminal company had to worry about the competition. If they tried to sell a 72 column terminal, and the competition had 80, they might lose out on some sales, even when the customers only needed 72. So, while 80 makes a lot of sense, matching card width, it isn't a direct path, and it might be hard to find a reliable source for the decisions of different CRT terminal companies, converging at 80. Gah4 (talk) 00:54, 1 July 2017 (UTC)


 * I'm not talking about the origins of 80-character punch cards, only that 80-character punch cards led to 80-character terminals -- that was the question at the start of this discussion. And, BTW, the story goes that punch card size was made to match 19th century paper currency size (8.5x3.5 inch), which was made to match 18th century bank-issued notes, which was made to match standard 17th century wax-sealed four-folded legal documents (8.5x14), which was made to match the size of the vellum used for royal and church documents of previous centuries, which was made from lambs for the finest skin, which yields a bit over 8.5x14 inches on average.  So, the origin of 8.5 inch cards dates back a few thousand years to the domestication of sheep.  But that's off-topic, and I wasn't involved in any of that.  I can tell say the first computer terminals came directly from card readers -- nothing else.  --A D Monroe III (talk) 02:52, 1 July 2017 (UTC)

This reminds me of the story tracing the size of the space shuttle solid rocket boosters to the width of roman chariots. It seems that it is convenient to standardize the chariot wheel spacing, as they then follow the ruts on the road from previous traffic. Not quite as obvious, this width lead to the rail spacing on railroads, though standardizing that wasn't quite that clear. And the space shuttle booster parts have to travel by rail. Following ruts in the road makes more sense then a standard printer width. (If you are just a little off, it will alternate with one wheel, and then the other, in the bottom of the rut.) Gah4 (talk) 00:54, 1 July 2017 (UTC)


 * Absolutely 80 column punch cards led to 80-column video terminals. With early electronics a power of 2 like 64 or 128 would have been chosen unless they had a really good reason for a harder number like 80.Spitzak (talk) 19:35, 21 November 2018 (UTC)

Mark Sense cards
There were certainly mark-sense cards that did not require a "special electrographic pencil". I think the image shows one. You needed a "number 2 pencil" although anything black would work. Also the cards were directly read by the same card reader that could read holes (not sure if there were two different light detectors or black underneath so holes looked like marks) and there was no equipment to punch holes where the marks were. This text is maybe describing the very first mark-sense cards and should be updated. Many people's only exposure to these cards were from ones they had to mark to program computers, and I certainly don't remember any "special pencil". Spitzak (talk) 19:40, 21 November 2018 (UTC)
 * In my youth I ran the main card reader at a university, and another chap did the read-sense cards for the school kids on his own machine. So, I'm pretty sure it was always necessary to read these with a special card reader. No doubt it would have been possible to have a dual-use reader, but I doubt this was common. - Snori (talk) 19:55, 21 November 2018 (UTC)
 * The one I am familiar with was attached to an HP2100A, and may have been an HP product. It probably dated from 1973 or later. It was not a high-speed card reader, I doubt it read more than 1/second. But it certainly read the "HP Basic" cards marked with pencil at the same time it read other punched cards (which were inserted in the deck to split each student's work apart and list and run them).Spitzak (talk) 20:00, 21 November 2018 (UTC)
 * True mark-sense cards definitely required a pencil that made conductive marks, as the marks were read by using a pair of brushes to try to pass current through the places where the marks would be. Early readers for punched cards worked similarly, by passing the card between brushes and a copper roller; where the brushes encountered a hole, they contacted the roller underneath, completing the circuit. (You can find diagrams of this setup in the IBM manuals for e.g. the 1402 card reader-punch.) This same reader arrangement obviously would not work for mark-sense cards!
 * The "special pencil" for mark-sense was simply a pencil that would make marks that were conductive enough, and it's true that some ordinary pencils could be used. Most inks wouldn't work, and #3 pencils were unreliable because they didn't have enough graphite in their "lead", and were too hard to lay down enough of a mark besides.
 * Later readers (for both punched cards and mark-sense) used optical sensing and then any mark that was dark enough would work. Still, the same reader mechanism could not be used for both types, as the marks you're supposed to make are not in the same places as they are on punched cards. Besides, punched card readers pass the card between the light source on one side of the card and a row of 12 or 80 photodetectors on the other. In this way the two can be close enough to each other that no lenses are required. All of the later card readers worked that way, including the very fast CDC 405 and the IBM 3504 (but not the earlier IBM 2540), right up until the end of the use of the medium. Since a "marked" mark-sense card would not pass light whether marked or not, it could obviously not be read this way.
 * For optical reading a mark-sense card has to be read by reflected light. Which is much more difficult and less reliable than reading a hole by transmitted light. In any case, per IBM, the term "mark sense" was only used for cards that were read electrically - though no doubt the term was used more generally by others. Jeh (talk) 10:24, 25 November 2018 (UTC)