Why were programs entered on punch cards instead of paper tapes?

Question

Dale Fisk's Programming With Punched Cards is a fascinating account of programming in the days of punch cards.

The fundamental dynamic was that early computers did not yet support timesharing.

The first attempt at allocating computer resources was simple queueing: each user got to go up and use the computer for a certain length of time. But this was inefficient: one manager complained that users were allocated 15-minute time slots, of which they would spend maybe ten minutes getting everything set up and only five minutes with the computer actually running their program, so the (at that time very expensive) computer effectively spent two thirds of its time idle.

Batch processing was a way around this; you no longer get to walk into the computer room; instead, you submit your program and data on punch cards along with everyone else; all the jobs are placed on a magnetic tape and fed to the computer as a single stream, so the computer gets to spend nearly all its time actually running programs; users collect their printouts the following day.

What I'm wondering is: why punch cards instead of paper tape?

It seems to me that the machinery to produce and consume paper tape was cheaper, being a standard fitting on ordinary teletypewriters.

Paper tape is also lighter, more compact, and doesn't have the problem of accidentally dropping a box of punch cards and getting the order scrambled.

It is less durable. Punch cards can last practically forever. But why pay for an archive-grade medium for a sneakernet job?

What am I missing?

Answer 1

The basic issue that paper tape is hard to edit. In theory you can cut the existing tape and splice in a new section, but in practice there is no easy way to find the correct location except by printing the contents of the tape (at 10 characters per second) and searching by hand. People did learn to read the tape hole patterns (they were no harder to learn than Morse code, after all) but even so, if a 1000-line Fortran program contained 50 differnent lines saying "I = 0" and just one of them should have been "I = 1", there was still the problem of finding the correct instance of what you were looking for.

Cutting and splicing the tape (using self-adhesive splices) required a certain amount of manual dexterity, and the joins were liable to come apart if the tape was used repeatedly.

On the other hand, punched cards would have their data printed on the card by the punch machine, so a deck of cards (stored in a card tray) could be read and searched as easily as a printed copy of the code, and individual cards removed and added as required.

Even in the cards-and-tape era, programmer time was expensive. The choice between making an edit in say 1 minute by swapping a few cards at "random" places in a 1000 or 2000-card deck, or maybe spending 15 or 30 minutes doing the same task on paper tape, was a no-brainer.

Having worked as a programmer in the punched card era, "dropping boxes of cards" simply didn't happen. Well, it did sometimes happen, but it was a rare enough event that the klutz who did it was going to be the butt of a ridicule for few weeks afterwards. In any case, cards were punched with serial numbers, so a dropped deck could be mechanically re-sorted.

In my university days, student computing access did use tape rather than cards, but most students' programs were small, and of course students' time was essentially free.

(And as a counter-example to the idea that tape didn't get "scrambled", I have seen situations where somebody was copying a tape after editing to get rid of all the splices and went to get a coffee while the teletype was doing its thing for 10 minutes. They came back to find the teletype had shaken itself off the table, and was busily destroying the remains of their tape while still running, upside down on the floor...)

Answer 2

Punch cards long long long predated paper tape.

But there's a practical consideration you're not thinking of. If you had ever used punch cards and paper tape, you'd know:

Punch cards can be dropped, and they scatter on the floor. Then you pick them up, put them all face up (by the printing on them) in the same orientation (with the one corner that's cut off lined up) and then you run them through a card sorter because you remembered to punch sequence numbers in the last 8 columns of the card! Then you're back in business, no sweat.

Paper tape can also be dropped. And then you've got a giant pile of spaghetti - spaghetti that's easily torn. Doesn't matter if it was spooled or fanfold: It gets tangled with itself immediately - possibly the only thing in the universe that happens faster than the speed of light. And untangling it is a slow process because if it didn't already get torn, it will when you untangle it (because the tape edges rub and catch against each other).

(And yes, as @alephzero has answered, you can't really edit it either, while you can always drop in another punch card to replace a bad one.)

P.S. - punch cards weren't just used for "sneakernet jobs". They did have to be durable because for a long time they were the only effective storage mechanism for programs. There'd be a bunch of separate rubber-banded piles on top of the card reader - one was your Fortran compiler, one was your assembler, one was the payroll system, one was accounts receivable, etc. etc. etc. A deck could be used daily and last months if not years.

Answer 3

Punch cards and paper tape are suited for different tasks.

1. Punch cards work better when the size of the data is not known in advance. How long would you manufacture the unpunched paper rolls? If the roll is too long, you are wasting paper and increasing costs. If the roll is too short, you either have to splice it together, or have the operator load more than one roll. In contrast, you use as many punch cards as you need for a job; extra cards can be used on a later job, with no waste.

2. Punch cards are superior for data entry. Many of the tasks of early computers were processing data. Tabulation of the 1890 U.S. census was done with punch cards created by Herman Hollerith. Businesses bought computers to process sales transactions, inventory, accounting, and payroll. Even some election systems as late as 2000 used punch cards for voting.

   What each of these applications have in common is they involve records with a moderate amount of data. Putting one record on each punch card works perfectly. With paper tape, you would be constantly appending records to the end of the tape.

3. You can insert or delete punch cards from a deck. This makes editing much easier. This was already pointed out in @alephzero's answer. Note that the ASCII code 0x7f was specifically chosen as the ASCII code for DELete because it was a way to rub out one character from punched tape, by punching all of the holes. It is also easier to edit a record by replacing a punched card; there is no simple way to do this with paper tape.

4. Punch cards are better for selectively using data. It is easier to pick out (manually or automatically) those records you want. For programming, you can re-use cards, as noted in @AlainD's answer.

5. When the data never changes, the above arguments go away, and then paper tape becomes superior. That's why player pianos had paper rolls, instead of punch cards. Certain programs such as loaders and assemblers were so stable that they could be distributed by paper tape.

Answer 4

To add one more dimension to the answers already given: cards are easier to handle, whether by operators, or in a cafeteria (user self-service) system.

A high-speed card reader has an input hopper and an output stacker. After reading a deck of cards, you've got a deck of cards in the stacker.

The high-speed paper tape readers of my acquaintance would do little more than spit the tape into a bin as it exited the read path. A medium-speed reader, 300 cps, is throwing out tape at about a yard a second. After it's all read, you have to roll it up again.

Answer 5

What I'm wondering is: why punch cards instead of paper tape?

Because it was already there?

Early commercial computers were made to replace tabulating machines. To do so they had of course to be able to read (and write) punch cards. There was no need for paper tapes.

---

Well, ok, paper tape was first, as the Zuse machines used them and some other experimental and scientific machines. Or not, as that was the time of primordial soup. Later on it was definitely only a thing for money strapped institutions. The real world revolves around business and business was all about card tabulating. That was were the money was the computer manufacturers were going for. That was were the computers were sold to. And that was were programs (and programmers) were needed.

So while there many good arguments can be made for punch cards (and I'd usually be first to do so) the real reason is way more mundane:

They used what was already available: Punch cards

It is never about what was invented first, but what environment a new technology is placed in. Existing data processing sites - the major early customers - had already everything needed for card handling, from keypunches over boxes and shelf for cards all the way to stand alone sorters and printers. That's everything needed to handle programming sans the compiler. Call it the IDE of the 1960s.

It as well helps to look at the contemporary landscape. Of the most successful computer line in history, the IBM/360, the most successful type was the plain Model 20. No floating point, no 32 bit, not even a full register set. By default delivered without tapes or disks (*1), but an all in one card processing unit, the 2560 Multi-Function Card Machine. It could handle the tasks of a reader, puncher, sorter and interpreter at once.

So, why would anyone introduce paper tapes to that world, making introduction of computers even more costly by requiring secondary investment in punch tape machinery?

---

Is that Sample Really Representative?

Users becoming time slots and setting up the computer on their own? Serious? It would be an extrem inefficient organization. I never have seen or heard of any computer shop set up like that.

Users normally never had direct contact to the computer. That was what the operator team was for. They were the guys - well, mostly gals - that carried the stacks of cards and paper around, feeding the machine to run with no slack between jobs.

Users would deliver their card stack together with a form describing what their input is / how it's supposed to be handled and what they expect as delivery of that job. This would be delivered to a counter outside the computing center - users never enter that place. It would be put in a box (if not already in one) and queued up with all other jobs.

Usually it was as well possible to send that job to be processed via in house mail (no, not Email, real mail, some guy pushing a cart), wich more often than not would also used to deliver output (and stack if that mark was set) back to each office.

In addition jobs could be stored at the computing center so only a form was to be mailed, stating which job is to be run at what data. Result would show up right at your office desk the next morning when the mail guy does his next round.

Plus them managing recurring jobs on their own. Just send a form (and a job) and mark it to be executed every day, week or month and you'll get the results without any further interaction delivered by mail.

Data processing is all about organization.

Computers just speed it up.

---

*1 - Which were only introduced later on and needed lots of RAM for system software, i.e. 8 KiB minimum :))

Answer 6

An other nice thing about punch card is that you can use them, or better use a small pack of them, in another program. Some sort of copy/pasting. I remember packs of hundrands cards with colored cards slipped from time to time to physically separate subprogrames that could be reused elsewhere.

Answer 7

There was paper tape and paper tape. For example, see Here for fan fold papertape. This could be read in by a conventional ASR33 type teletype reader but often was read by a specialized high speed reader. The fan folded tape could be fitted on one side and read through quickly. It could then be rewound almost like a tape from one side to the other. The first shows a diagnostic, but one of the early control systems for the Dartford Tunnel in the UK used something like this for a complete O/S (RSX-11S) with applications to control the signals, lights and fans. The tape in this case was made of mylar rather than paper which tends to be more robust. This would be read when the controlling system (a PDP-11) needed to reboot. The point being that this was binary rather than text so there was no need to edit at the line level.

Answer 8

Another advantage of cards over tape: Ease of setup for batch processing:

• With cards, it is a trivial effort to pile up multiple stacks for successive scanning.

• With tape, either you have to splice multiple tapes together in sequence and respool (a process fraught with peril of tearing), or dismount each tape and mount the next, which takes up a fair amount of costly time.

Answer 9

A few additional side benefits of punched cards as opposed to punched paper tape:

• Punched tape was supplied on rolls of particular lengths. A roll of tape could only accommodate a certain amount of data. A stack of cards could be any length.
• Paper tape wasn't as robust as punched cards - the paper was thinner than that used for cards. Punched cards were read by placing the card against a shiny metal backing plate and shining a bright light so the holes in the card could be read. Punched tape was read by pins pressing against the tape and passing through the holes. Over time, this mechanical action could wear out the tape.
• A crease in a punched tape would be problematic for the reading device.

Answer 10

Other reasons:

• Economics. Making a long roll is more expensive than cards. Shipping cubes is more space efficient than cylinders. If one section of the tape is damaged, it can't be sent to customer, but if a few cards are damaged, such as with oil from cutting machine, just replace them with non-damaged cards.
• Coding repair. If you make a coding error on one card, then all the others do not need to be replaced to fix the problem.
• Structure: You can make a card stiffer, and it will resist damage and wear more than the paper tape. If you make the tape thicker , then it breaks down faster and you get less length per diameter of roll.
• Ergonomics: Human handle hand-sized things more efficiently. If it is substantially smaller or larger then it is less easy to use and more easy to break/harm.
• Adjacent markets improve economy of scale. Using the same stuff as 3x5 cards, birthday cards, signs, or such means you can have a larger market for the same operational equipment and same fundamental process. People don't use paper tape rolls for birthday cards, or study notes.
• (repeat) Modularity: the tape is a single unit while each card is its own unit.
• standardization: it is easier to standardize the system on the card than the tape, and once standardized it allows better economy of scale
• many sensors prefer to operate on a planar surface than on a cylindrical one.

Answer 11

As pointed out in the other answers, trying to "edit" paper tape directly is awkward.

One solution is to have a batch editor that reads an edit tape into memory, then use the batch edit commands to read a source tape and punch an edited tape, according to the edit commands. A second tape reader could be used to reduce memory requirements.

A more common situation is to keep text files on another medium other than paper tape, such as a disk or magnetic tape. Then paper tape is only used for the initial entry of data for text file, or for batch edit commands. This is what was done at my first job in 1973.

As for long term backup, that company I worked for in 1973 would rent a high speed punch (400 characters per second for paper tape, 100 characters per second for mylar tape) and punch all of our source files onto mylar tape. We also backed up to magnetic tape, but if I recall correctly, mylar tape has a lifespan over 100 years.

As for paper tape getting tangled, at that company in 1973, we used a 600 character per second photo reader that just spit out tape at 5 feet per second into a pyramid shaped pile onto the floor, and we used high speed electric rewinders to wind the paper tapes back into rolls, by winding the tape around a circular set of pins protruding from a flat round disc for easy removal of rolls after winding. Tangles and/or snags were very rare, and even when it happened, it rarely damaged the paper tape, unless the paper tape was really old. Dealing with old paper tape was not an issue once we started doing backups to mylar tape for long term (over 100 years) storage. If paper tape was torn, we had laced perforated scotch tape used to splice paper tape. The splice would not impact usage by teletype or high speed photo-electric reader.

In the case of the HP 2100 series of mini-computers, the delete code deleted the entire current record rather than just a single character, so if a programmer made a mistake during entry on a teletype, delete, carriage return and line feed were used to delete that line and the line was re-entered. This was just a convention used by the HP mini-computers. I don't know if this convention was used on other systems.

As a bit of trivia, Bill Gates and Paul Allen developed Basic for the Altair 8800 system (Intel 8080), bringing paper tape with the compiled output to load into the Altair 8800, only to realize on the plane that they didn't bring the load program, so Paul coded up a load program during the flight to toggle into the front panel of the Altair 8800.

Answer 12

Back in 1962, when I was in a class to learn machine "coding" on a Burrough's 205A, the first 6 instructions were hand entered with toggle switches on the actual computer rack. They were the "start up" instructions that made the 8 bit computer smart enough to read the paper tape accessory on the side of a KSR-33 teletype. The 205A had only 2K or rotating drum memory and was all electron tubes (12BH7's as I recall).

Paper tape was miserable because it was unforgiving when you made a mistake. This was before electric typewriters. Every instruction was a single line with a fixed memory address followed by the three hex character instruction and the object address. I.E. Fixed width instruction fields

No "micro code". No IEEE algorithms. You wrote your own subtraction routines as two's "complement add". Multiplication was shift and add. Division was a nightmare and fractions were an adventure as was decimal to hex (no octal for you!).

Hollerith cards, card punches, and sorters arrived with the IBM machines. The terror there was having errors of "0" or "O" (note the zero's were not slashed on the card punches) mistyped and buried in the deck.

All in all, I remember it as a good time and exciting to be at an edge of technology.

(EDIT) After 55 years, I wrote it wrong. What I see in the 205 Manuals is not what my mind remembers. I visualized octal and not BCD and 11 digits, but that could have been a PDP-8 remembrance. . . from a few years later.

If you are curious about the instruction set, it is online at:

http://bitsavers.org/pdf/burroughs/electrodata/205/3021_B205_Central_Computer_Hbk_Mar56.pdf

The startup procedure is on page 3-4 at :

http://bitsavers.org/pdf/burroughs/electrodata/205/3044-A_B205_Operating_Procedures_Jun60.pdf

The control console is more what I visualize.

The Burrough's manual notes that there was a thermal sensor on the exhaust that shut the unit off after the exhaust temp rose above a preset number for 15 minutes. HVAC maintenance was more important than the electronic technicians.

As students, we were given lessons about register arithmetic and number base conversions between binary and BCD, word processing was unknown."Sort-Merge"came much later as did a higher level language that was compiled.

In: Santa Cruz Sentinel, Volume 106, Number 44, 21 February 1962 Issue Page 34

This is a picture of a few of the high school class in front of the 205. Always wondered how many of us went on to careers in electronics or CS.

Answer 13

I'm no historian, and this is not a valid answer. It is just that it cannot fix in a comment. But I don't see why punched card any better than fanfold tape.

• Tape predates cards. Folded tape dates back to the 1920s(US1541201A).
• Colossus tape reader reads at 5000 cps in the 1940s, doubled card readers.
• Folded cards don't need rewinding or waste packaging space.
• You have to wait and refeed stacks of cards. Not the case of folded tape.
• You shouldn't expect non-tangling if you don't buy stiffer tapes.
• Every fold can be regarded as a record as cards do. But cards are no stream.
• If you can print line number at the end of card, so can each fold of tape.
• With line numbers, a buffered reader doesn't care if lines inserted or appended.
• Looking through folds of tapes is as easy as cards, and inserting not too difficult.
• Thiner tape wastes less paper even if you leave half fold of tape blank just as card.
• Paper tape can be full of holes while lace cards damage card reader.
• If binary mnemonic is needed, even narrower tape can afford some.

If these reasons still don't stand, why there is no narrower, binary, separate but splicable cards comming into market?