News for hackers. Stuff that matters.
Re:Virtual IBM Punch Card Data Processing by khaledh on Tuesday August 15, 2023 @06:56PM (#37140376)
I highly recommend watching this YouTube channel[1]. It's the ultimate IBM 1401 simulator running on the Unreal engine. The level of detail is mind blowing.

[1] https://youtube.com/@rolffsonsIBM1401channel

Re:Virtual IBM Punch Card Data Processing by nxobject on Wednesday August 16, 2023 @04:53AM (#37144399)
If I had the money, I would buy a VR headset for this developer in a heartbeat. It would be fascinating to see what that would add the experience. Especially if he added all of the sounds, to get that full HVAC experience...
Re:Virtual IBM Punch Card Data Processing by tempodox on Wednesday August 16, 2023 @12:21PM (#37149412)
Here is the machine in action:

https://www.youtube.com/watch?v=GZtLMMuLUuY

Downloading the simulator is mandatory, since we're saving over $10 per minute in today's money compared with renting the original hardware. Also, the signs say “THANKS” instead of “THINK!” which is a nice touch :)

Re:Virtual IBM Punch Card Data Processing by AnimalMuppet on Wednesday August 16, 2023 @01:13PM (#37150207)
Wow. 16,000 characters of memory (max), 11.5 microseconds cycle time, so 0.087 MHz = 87 kHz frequency.

Wikipedia says $25k/month rental in today's dollars, which I calculate as being $34/hour. I could buy that much computer today with the loose change under my car seats.

Re:Virtual IBM Punch Card Data Processing by tempodox on Wednesday August 16, 2023 @02:11PM (#37151208)
Indeed. The supercomputer in my pocket does high-resolution domain coloring and fractal generation in seconds or even just a fraction thereof, while the 1401 could probably have heated a house for a winter with the same compute load.
Re:Virtual IBM Punch Card Data Processing by cogogo on Tuesday August 15, 2023 @06:59PM (#37140401)
Started my career as an IBM SE for the mainframe in the early 2000s. They were desperate to bring young people into the mainframe world to carry the baton. My peers were all guys with kids my age or well older and most of them carried punch cards around in their breast pocket to take notes. At that time they were still abundant in random IBM supply closets across the midwest.
Re:Virtual IBM Punch Card Data Processing by themadturk on Tuesday August 15, 2023 @07:22PM (#37140647)
When I attended Washington State University in the mid-1970s, my girlfriend was a chemical engineering student taking her one and only required programming class. I remember helping her punch cards late at night at the computer center, and used "liberated" blank punchcards as bookmarks for years afterward.
Re:Virtual IBM Punch Card Data Processing by vincent-manis on Tuesday August 15, 2023 @07:40PM (#37140839)
Wow, there's also a virtual Honeywell 2000 computer. I have not-so-fond memories of finding a Honeywell 200 system in my university's administration building. It was the only machine on campus with a Cobol compiler. Compilations took 30 minutes or more; the admin folks kindly allowed me an hour at a time on the machine. I don't remember much of the system now, though I do remember that Pass 2 of the compiler was the system sort utility (don't ask me why). I believe that the 2000 was the Big Sibling of the 200.

Also, on the subject of punched cards, Datamation ran an article in the 1960s that dealt with the nasty subject of storing blank cards (every computer center had a room with huge boxes of punched cards and of printer paper). You could get rid of your card storage problem by running a cool program. Whenever you need a box of cards, you run the program, put a box of blank cards in the punch input hopper, and presto! blank cards in the output. Believe it or not, Datamation claimed that they'd had a number of requests for this magic program.

Re:Virtual IBM Punch Card Data Processing by ok123456 on Tuesday August 15, 2023 @07:57PM (#37140958)
Reminds me of the old x026 program that simulates creating punch cards.
Re:Virtual IBM Punch Card Data Processing by 082349872349872 on Tuesday August 15, 2023 @08:46PM (#37141341)
Re:Virtual IBM Punch Card Data Processing by masswerk on Wednesday August 16, 2023 @01:07PM (#37150114)
I always wanted to implement at least FORTRAN (VI or maybe 77 with backward compatibility) for a period correct language, but there are always other things to do…

(Having said that, using a familiar language like JS may be suited better for experiencing the hassles of the process, anyway.)

Re:Virtual IBM Punch Card Data Processing by Animats on Tuesday August 15, 2023 @10:49PM (#37142324)
That article shows the machines, but gives no sense of how they are used together.

Here's a simple billing cycle. Inputs are:

- Purchases, with a description, customer number and amount. These were created manually, by a keypunch operator. Cards would also have date and card type, copied from card to card while punching.

- Payments, with a customer number and amount. Again, with date and card type, copied from card to card.

- Customer addresses, usually a 3-card set, with name and address, customer number, and card type. These are semi-permanent, used month after month, updated by hand.

- A previous balance card for each customer, with customer number, amount, date, and card type.

So, at the end of a billing period, the first step is to sort all the purchase cards by ascending customer number. That's what the 082 sorter is for. They should have been accumulated in date order, although if you have to sort by date, you can. Sorting is labor-intensive, because all the sorter does is a 1-digit radix sort. So if you have a 5-digit customer number, you have to put all the cards through five times. You don't want to have an unnecessarily long customer number. There's a lot of manual card handling here, but it's mindless.

Payment cards get the same sorting treatment, by customer number.

Previous balance cards and address cards should already be in order by customer number, and that will be checked.

Now the 085 collator is used. First, the address cards and previous balance cards go into the two hoppers. The collator can read selectively from either hopper. The only operation the collator can do is compare numbers, less, equal, or greater. But that's enough. The collator combines the address deck and the previous balance deck, so that the deck, for each account number is now

    Address 1, address 2, address 3, previous balance.
Any previous balance card without an address or an address without a previous balance card is caught at this time. Those cards may go into a different output stacker, or the collator may be stopped for a manual check, depending on how the plugboard was wired.

Next the another collator pass - the deck from the previous step vs. the payment cards. Now the deck is

    Address 1, address 2, address 3, previous balance, payments.
Now the big collator pass, where the purchase cards are merged in.

    Address 1, address 2, address 3, previous balance, payments, purchases.
The collator has checked that everything is in the right sequence, with all cards in a block having the same customer number and successive blocks having ascending customer numbers. That's the final deck, and it now goes to the 403 tabulator.

The tabulator has one hopper and one stacker, a printer, and a big removable plugboard that's the program. The data processing shop will have a rack of such plugboards for different jobs. There will be custom-printed multi-part carbon forms, and a "carriage tape" loop, which tells the printer how much to advance the paper to get to the next section of the form. The 403 has very little memory, about 40 decimal digits, although you can buy a few extra counters from IBM if you have to.

So now it's time to print the bills. The final deck is loaded, 500 cards at a time or so. The 403 grinds into action. The address cards are read, and they print in the address block on the form, which will be visible in a window envelope when mailed. Totals are cleared. When the tabulator detects the change from an address card type to a previous balance card, the carriage is told to advance to the line indicated by a hole in the carriage tape. This moves the paper to the beginning of the transactions area. The balance forward is printed on the balance forward line, and added to the total. The purchases follow, again being added. The payments follow, and those are subtracted. When the reader sees a new address card, that indicates the end of this customer.

So the tabulator advances to the total line on the form, and prints the total amount due. An invoice has been generated, and will later go into an envelope to be sent to the customer.

Remember that 514 Reproducing Punch? That's been cabled to the 403 tabulator via a huge cable and connector, for Summary Punch mode. When the tabulator prints the total amount due, it sends the total and customer number to the 514, which punches a previous balance card with the new balance, to be used next month.

You get the idea. One cute feature is that you can program the 403 to detect a hole punched in the carriage tape which indicates the form is full. Then it can print a per-page total, increment the page number, start a new form, repeat the customer number, and continue printing transactions. Yes, they thought of that problem.

Re:Virtual IBM Punch Card Data Processing by whisket on Wednesday August 16, 2023 @06:12PM (#37154580)
Thanks so much for writing this out.

It's lovely to see the physicalization of arrays as decks, fixed-function units as sorters and collators, ALUs as tabulators, and human beings as data sequencers.

I really appreciate you sharing this.

Re:Virtual IBM Punch Card Data Processing by Animats on Thursday August 17, 2023 @02:02AM (#37157813)
This is batch processing. Many systems today, especially ones that process money, still look much like that. The hardware is different but the concept remains. Sequential files are created, sorted, merged, and outputs generated.

It's a reliable, well-understood process. If something funny happens, somewhere there is a logged transaction that made it happen. You don't just have an inexplicable database state. Auditors look for those transactions that made something funny happen. What you see in online banking is often yesterday's batch run plus interim updates during the day. The real updates happen in some batch job.

Re:Virtual IBM Punch Card Data Processing by euroderf on Wednesday August 16, 2023 @01:08PM (#37150124)
At college in the late 70s, the breakthru was a a new room full of terminals where you could make "virtual punch cards" (yes) for the mainframe. The name of the terminal program was Roscoe - pretty much un-PC.

These virtual punch cards sure were an improvement from the tangible punch cards. The turnaround for all that "SYSIN DD *" JCL and FORTRAN and PL/1 came down to PDQ.

It was almost like you could feel the earth move under your feet... from batch execution to almost-interactive execution.

Re:Virtual IBM Punch Card Data Processing by zabzonk on Wednesday August 16, 2023 @02:08PM (#37151169)
i wrote something on an ibm pc back in the early 80s to replace several univac card punches, which we could not get sensible maintenance contracts for anymore. the program allowed the operator to specify tab-stops (much like a typewriter) and specify if a field was alfa or numeric. these were being used by some of the most fiendish (and in many cases the pritiest) data input ladiews i have ever come across. the input was written on to an msdos disk file, and then transferred to mainframes and minis via kermit.