Why Build Colossus? (Bill Tutte) - Computerphile
Vložit
- čas přidán 13. 09. 2018
- Cracking the code was only half the battle. To keep the upper hand, when using Bill Tutte's statistical methods, the detailed counting had to be automated - enter Colossus! Professor Brailsford takes up the story.
The professor's notes:
www.eprg.org/computerphile/lor...
Bletchley Park Playlist: • Bletchley Park (Coloss...
Professor Brailsford used the book "Colossus" by B. Jack Copeland and others (Oxford University Press, 2006). Also recommended are chapters 18 and 19 from : "The Bletchley Park Codebreakers" by R. Erskine and M. Smith (eds.) Biteback Publishing 2011
/ computerphile
/ computer_phile
This video was filmed and edited by Sean Riley.
Computer Science at the University of Nottingham: bit.ly/nottscomputer
Computerphile is a sister project to Brady Haran's Numberphile. More at www.bradyharan.com
I'd listen to Brailsford talking about pretty much anything. Keep him on as long as possible :)
Lol. I watch more of his videos than all other Computerphiles combined. Probably re-watch more of his videos than others
Awesome video as always. From the video, "Tommy said to them, I've been doing research on use of thermionic valves in telephone exchanges..." Valves had been used for since 1915 to amplify long distance calls. The genius of Tommy Flowers was his research into _switching_ calls electronically. He was way ahead of his time... and sadly under-recognized for his contributions.
Also, I saw elsewhere that Newman turned down the fully electronic idea at first, but Tommy Flowers went back to Dollis Hill and built it anyway. Then later when they realised he was right he delivered it within days. So sad he died poor because he spent his own money on building it and nobody thought to at least pay his expenses. Disgusting really, and he just went back to his job at Dollis Hill, someone like Newman could have helped him but there was a lot of snobbery then and Flowers came from the east end and sounded cockney. Sigh.
*recogniSed
I always love listening to him, doesn't matter what the topic is. Such a treasure.
I love all the videos with professor Brailsford!
Yeah, he seems like a genuinely nice guy with a lot of knowledge to share, and the enthusiasm to share it well. I'd watch more videos with him any time. :)
Look how times have changed: from "Never turn off the computer" to "Have you tried turning it off and on again?"
I still live by the mantra of never turning computers off unless hardware needs to be added.
Years ago I noticed that most of the PSU and harddisk failures would occur after power-downs,equally for planned and unplanned outages. Regardless whether the servers were brought down for a soft landing via UPS's. After startup, there would almost always be something that died.
What about power consumption? Unless a secondary goal is heat, you're wasting electricity.
DJoppiesaus? What about it? It is mine to waste!
Also keep in mine that the energy expenditure in creating a new computer, the case, components, everything, takes so much more energy than an average single system will ever use in it's lifetime.
Keeping hardware alive for longer actually saves energy on the whole.
Only on Windows
"Have you tried turning it off very slowly and then turning it on again very slowly?"
Professor Brailsford is the boss!
I love the WWII and code breaker talk. Please keep him talking about this.
Such an interesting topic and such a calm and intelligent man
Legend...
And Tommy Flowers, being a practical man, not just theoretical, was spot on... Valve last for years if you just leave them running.
the english being all like "you know that encryption algorithm you thought was beyond human capability to decipher? well, we cracked it" was one of the biggest mic drops in history
I can listen to this guy forever
This is one thing assembly language programmers usually learn. Back in the 8 bit microprocessor days, you used to look at how many bytes an immediate load used to take, like LD A, 0 on the Z80, and how long it took to execute, and compare that with a similar operation like XOR A, A, a lot of times it took less time (clock cycles) or less space or both to do the XOR. As long as you knew this as an idiom, or made code comments, it didn't reduce readability/intelligibility.
I use valves (tubes in my case) all the time. They are still quite common in medium to high power transmitters. The same holds true, we like to bring the filament up slowly using a variac or even a built in so called soft start circuit. They are highly reliable if treated properly and can be abused much more so then transistors in similar applications. As the old saying goes a MOSFET is one cycle from exploding at all times! Don’t get me wrong I don’t want to go back but it amazes me how fragile people thought tubes could be and my experience is exactly the opposite
His heat sync speech on tubes is still valid today on all electronics or breaker boxes.
Thanks again Professor Brailsford
My uncle worked at Dollis Hill, with Tommy Flowers.
He never spoke of his work. (Wretched man!)
My father worked there as well before the war but never talked about it much and he left in 1938 to join the BBC.
I Love all of computerphile videos
It is really funny that I saw the your Enigma Videos about a year ago, decided to make it the topic of my seminar paper (the exact Topic is 'deciphering machines at Bletchley park') and just when i came to the chapter about the Tunny and Colossus I found this video uploaded not long ago xD
Dr. Forbin's answer was quite sufficient, thankyouverymuch.
heh, I have something to attest to the power of tubes being gently heated and kept at slow-ish smoulder...
The Hammond Novachord :3
It took all that to decipher messages from the enigma machine, and a modern office computer could perform every calculation colossus did in it's useful life in a fraction of a second
I love listening to him.
Is there anything better than a Brailsford video?
Double Z is actually rather rare in German, since for most words TZ instread of ZZ is used.
Interestingly, in all the books I have read about Alan Turing, Bletchley Park and so on, there is very little connection mentioned between Alan Turing and Tommy Flowers, and I believe Turing had long moved on to Hanslope Park (early 1943) to work on voice encryption by the time Colossus actually became "Turing Complete" and properly programmable with decision branches and loops. So we end up with the man who defined what a computer is, and the "first" (yes, Konrad Zuse etc, I know) programmable electronic computer in the same place, but with seemingly no link between them.
Much of the information about Colossus remained classified WELL after Enigma, well into the mid-2000's. The blue book on the table under his papers is "Colossus: The secrets of Bletchley Park's code-breaking computers" by Jack Copeland, worth a read but it is a long book. More information is coming out now that authors have keywords to do their FOIA requests or whatever the British equivalent is. The Imitation Game staring Benedictory Cucumberpatch stirred up a lot of renewed interest in Turing and others from that time and Bletchley Park in general.
@@dustysparks Some of it is still secret I read, probably for another 29 years, like other records from WW2. I would love it if Cucumberpatch would play Tommy Flowers but that would be too confusing and he's too posh. Maybe "My name is ..." or someone else a bit younger. I am reading Copeland's book and have a personal connection to Dollis Hill as my father worked there before the war.
Because building Colossus increases trade?
Fittingly, Elizabeth even just beat me to building it. :-|
I could never wrap my head around that code could be cracked by statistics. but now i can thanks to him!
Imagine the change that would have happened if someone took a simple Raspberry Pi back in time to the scientists/arithmeticians working on this stuff.
I doubt they had the necessary ports back then.
I've just realized why I love hearing Professor Brailsford speak - he sounds just like Winnie the Pooh!
Oh bother
I like this guy
Brilliant!
The ABC (Atanasoff-Berry Computer) always gets the short end of the stick, I think. It preceded the Colossus and was the first computer to use Binary. It also relied on vacuum tubes for the computation.
Yes but it was hard-wired to do one particular problem, Gaussian elimination with a set of equation coefficients. And the memory was mechanical. Close but not quite what we consider a fully electronic computer, and definitely not programmable.
This is similar to X-Ray sources for crystallography (Cu, Mo, etc.)
There is a working-voltage and a standby-voltage and these instruments (to be used by researchers) is only to be turned off for maintenance or long pauses in work schedules
could just have droped the double leters in their mesages
Jiyva I see what you did there :)
Wow. I mean they could've simply prohibited double characters. You would be able to properly understand the whole message once decrypted but would get rid of this weakness by the cost of an order...
Wouldn't there still be a problem of coded numbers for things like coordinates and quantities? I'm just guessing.
Yes there were rules about repetitions and a suggestion to insert plenty of nulls. Not followed?
Always love a strategically-placed Rubik's Cube
*_...so you're saying Germany lost their war because they didn't convert all-double-letters to singles, like ss = ß... you'd probably also have to figure all-statistically-interesting-pairs that produce more 0's, (even depending on various, shiftings), if, you know, their set of character-to-bit conversions..._*
I don't understand that with the double letters. The plain text is XORed with the key text, but double letters would only influence the balance between 1s and 0s if you'd XOR adjacent letters within the plain text according how it is described in the video.
Yeah, me too. Why would you XOR adjacent letters? No idea.
Because they are good against light units
I don't know why the Colossus computers didn't use sub-miniature valves as the computer would've been smaller and used less power.
There were only a few miniature valves back then, the 6J6 and 6AK5, and neither one was available in the UK. There were a few subminiature tubes but their whole production was used in making VT fuzes and walkie-talkies. Also the subminiature tubes were not very reliable, they were very fragile, except for the VT fuze ones.
I wonder what the clock speed was for that thing. A few Hertz? 100?
Daniel - 25 kHz, limited by the maximum speed of the input paper tape rather than the electronics.
Rather than trying to synchronize the tape drive speed to the computer’s clock, the clock was based on a photocell for the tape’s perforation, so the clock speed depended entirely on how fast the tape was going past the photocells, which was typically 5000 symbols per second. So the clock speed was 5kHz. There were 5 parallel counting units, giving 25,000 operations per second. Overclocking was possible, but tended to break the tape.
would the Germans have known character doubling was an inherent flaw in encryption? perhaps a "drop one letter from all double letter words" directive should have been implemented!
Strange thing is Flowers was never allowed to build his electronic telephone phone exchange
@0:04
What? So if they "transmited" the "mesage" with purpose "mispeling" by removing the double chars... the "mesage" would be readable but not crackable?
Great thought! Also check out my comment above about how you could design an encoding scheme to balance out the 1's and 0's. With your method and mine I feel like the resulting decryption task would be MUCH more difficult.
There are some simple ciphers that disallow double letters by inserting a character in between, like X. This isn't for security, it's because they encode letter pairs and will choke on a double letter. I don't know if they add the X when the doubled letter falls in two separate letter pairs, but regardless, this increases the frequency of the letter X which itself can become a clue for cryptanalysis.
In other words, I think changing the clear in a systematic way would end up generating a statistical bias. It would just be a _different_ statistical bias from the one it was trying to solve.
You shouldn't, they get easily countered by the zerg's corruptors.
Or vikings...
Interesting!
So how did they go about estimating which letter the exclusive ORs represented? A neophyte here, thanks a lot!
The frequency distribution of double letters in most languages is fairly easy to determine in advance. With German, it could be something as simple as the old fashioned character ß (long S) being transmitted as ss on the teletype machine.
Why? To play Numberwang, of course.
We should never have insisted on 'stickstofffreie Schifffahrtsspeziallagerrechte'!
How did the British secure their empire wide communications?
Cool Hawaiian shirt!
For our yank friends, thermionic valve is a vacuum tube.
Any one watching these types of videos is hip to that, LOL.
And now... you could probably do it in a crappy Arduino, let alone a Raspberry!
of course, those machines are a billion times faster than Collosus .
If only the Germans had been broadcasting in lipograms excluding double letters, lol.
... To crack the code... Duh...
His nasal voice reminds me of winnie the pooh... 😅
Oh go on.... you're recreating the 40s.... say "stroke" not "forward slash". ;)
Pizza! 2 Z
Or pizzazz. I wonder if the military every use that word though.
1st
damn, 2nd here ;)