Slashdot Mirror
Electronic Abacus
yoey writes: "Blast from the past in an article at the Economist: There are those who do not believe in the desirability of introducing anything as esoteric as electronics into business routine at all. Others believe that there is a limited field for electronic methods, provided that they fit into, and do not disrupt, established business systems. But there is a third group ... who consider that a major revolution in office methods may be possible. This revolution would involve scrapping the greater part of the established punch card calculating routine and substituting a single 'electronic office' where the giant computor [sic] would perform internally all the calculations needed for a whole series of book-keeping operations, printing the final answer in and on whatever form was required."
i've often heard about the LEO (Lyon's Electronic Office) being for the first business application of computing but have yet to see anything about the tech they used and how they developed the system. can anyone point me out to some online info?
In the good old days, "computer" was a person who computed, so another word had to be used for a machine that calculated things. I had a professor who was in charge of a WW2 lab of "computers" who were mostly attractive young women. He was of the opinion that things had gone downhill in the computing field since then.
So even if the original author used "computor" to indiciate some different meaning or usage, but a large part of slashdot would assume it was a typo (which we evidently would), [sic] is appropriate.
Man, this is an unusually anal post for me. It's too cold.
-Puk
p.s. For what it's worth, webster doesn't have "computor".
Where this article talks about "scrapping the greater part of the established punch card calculating routine", it isn't about scrapping the punch card keyboarding machines, but about using the cards as input instead of as the databases themselves. From the 1880 census until the 1950's large databases consisted of boxes (and sometimes cabinets or rooms) full of punch cards. To process the data, there were a number of specialized machines:
Keypunches: a keyboard that punched holes into cards. Good ones also typed the data along the top of the card so it was human-readable, and could copy part or all of a card. The ones I worked with could be programmed by typing control codes onto a card and wrapping it around a spool inside the machine -- this gave you tab stops and let you set it to automatically copy headers from each card to the next one, until you hit an escape key to let you change the headers...
Sorters & mergers: Sorting and grouping was accomplished by machines that would physically shuffle the cards into order. The operation was counter-intuitive; to alphabetize a 20 letter name field, you'd start by sorting into 26 bins on the _last_ (rightmost) letter, stack them up and sort on the next letter (which left cards differing only in the last letter in order), and repeat for 20 times through. Searches were done by setting the sorter to set aside cards matching the criteria and running the whole set through. And after doing a search or other operation that split the deck into two categories, it was nice to have a machine to merge the two decks back together into order without requiring a full-scale sort.
Tabulators: Would read the sorted, grouped, cards and add up the columns. Also could perform calculations on a card (like hoursworked * payrate = grosspay) and punch the answer into the card, or onto a new card. Tabulators generally did not type human-readable text on the cards, so...
Printers: One kind would read cards and type text along the top. Some of these were still in use in the 1980's, because mainframes still could output to card punches, and those punches did not type text... The other kind read cards and printed the report on paper.
When I started hanging out at the college computer center (1971), the databases were kept on removable hard disk packs, and punch cards were mainly for data input. However, even though they'd keep 3 copies of each database on different disks, the reliability was low enough that for really important stuff they'd also store the punch cards as a backup, or sometimes have the computer punch a backup into cards. The machine that printed on those cards was kept running, just in case. At least a half-dozen keypunches were in continuous use (and the card reader on the computer had to be overhauled once a week so it could continue reading all those cards). The tabulator was just gathering dust, but the sorter was used frequently -- batch database updates run faster if the input is in the same order as the disk file.
For a wealth of information on the computer mentioned in the article, the LEO, see:
www.leo-computers.org.uk [i.hate.square.brackets] [probably.already.slashdotted.to.hell]
What you have to realise is at the time, my Dad and other people working on the LEOs genuinely believed that these were the world's first computers ever, not just the world's first business computers as they later became known.
You see, at the time, all the World War Two computer developments were covered under the millitary Official Secrets Act.
When these secrets broke to the general public in the 1970s, needless to say my Dad was somewhat dissapointed to discover he was not a great computing pioneer after all!
My Dad fondly recalls being able to boil a kettle and fry bacon & eggs on these monsters.
Andrew Oakley - www.aoakley.com
This is all documented in the book "the Chip" by T. R. Reid, which I literally have on my desk as I write this. It is briefly summarized here:
In other words, it wasn't just solid state as in a single transistor, but solid state, as in entire cirsuits, the integrated IC that was the solution.The problem was that transistors still had to be interconnected to form electronic circuits, and hand-soldering thousands of components to thousands of bits of wire was expensive and time-consuming. It was also unreliable; every soldered joint was a potential source of trouble. The challenge was to find cost-effective, reliable ways of producing these components and interconnecting them.
The Tyranny of Numbers was quite real, and occupied minds for most of the 50s. The solution of this basic and fundamental problem made possible the computer age. They are probably as important as the binary logic that runs on them.
You can also read more about this on the Texas Instrument Website.
"It is a greater offense to steal men's labor, than their clothes"
The operation was counter-intuitive; to alphabetize a 20 letter name field, you'd start by sorting into 26 bins on the _last_ (rightmost) letter, stack them up and sort on the next letter (which left cards differing only in the last letter in order), and repeat for 20 times through.
This is the radix sort, which all hackers should learn while teething. When you apply certain rules that ensure that only columns that might be significant are compared, it is an extremely efficient sorting algorithm.
Worst-case, the complexity is O(l n), where l is the length of the longest string and n is the number of strings. With completely random data, l is effectively log n, so overall it goes to O(n log n). The extra rules substantially reduce the effect of lack of randomness in strings, so it's likely that the algorithm will almost always run in O(n log n)
Compare to a merge sort, which is O(n log n) worst-case, the best you can get, but that assumes that the comparison step is constant. With a string, worst-case comparison is O(l), resulting in overall performance of O(l n log n) or, with random data, O(n log^2 n). QuickSort is even worse, with a worst-case performance of O(n^2 log n), though still an average performance like the merge sort. (Too bad I can't use superscript on this board.)