Lorenz, Colossus and the Dream of a Universal Machine for Cryptanalysis

A talk by Andy Clark of TNMOC and Royal Holloway at the "ENIGMA - Precursor of the Digital Civilization" conference organized by the Foundation for Polish-German Cooperation on 10 November 2015 in Warsaw. (fwpn.org.pl/en/ )

[@johnballantyne3231]

What a wonderful presentation. So glad to know that, at last, those brilliant wartime participants are now being acknowledged world wide. Just wonderful!!!

[@douglasdrummond790]

My first experience with an electronic computer was the Bendix G-15D, in 1963. This machine was very closely based on Turing's ACE, using 450 electronic tubes/valves. After visiting Blechley Park in August 2014, I knew I *had* to write a simulator program for the G-15 which was already on my list of antique computers to simulate. I am about halfway through debugging the G15 simulator. I had already written a simulator for another antique machine, the Royal Typewriter Company's LGP-21. I am a retired computer & electronics engineer with 52 years of programming.

[@casparharte]

Is it true that ACE ran at 1Ghz?

[@briancase6180]

@@casparharte yes, it's true, but it needed one million cycles to execute a single instruction.

[@granskare]

the 3 Polish guys who first solved codes are real heroes.

[@RudneiDiasdaCunha]

Thank you for providing this wonderful talk!

[@jpkiwigeek]

This was truly fascinating, thank you very much. Mr Clark speaks of earlier seminars during the day relating the earlier code breaking schemes - was this also recorded? Is it possible to see these as well?

[@myparceltape1169]

That was fun to hear how much faster the more up to date method was. Not only WRENs read Morse, I can think of a WRAF. But you are right, many service people sent in transmits from round the globe.

[@TheFleetz]

Excellent presentation! Thank you...

[@MrDaiseymay]

MARVELOUS, CONCISE CLARITY.

[@TheHornoxx]

Prima - was für ein spannender und guter Beitrag über so viele gute und kluge Köpfe 👍

[@amrkoptan4041]

excellent video

[@samuelelsby1800]

Are the previous talks of this conference referred to, available?

[@tnmoc]

There are some other talks on our KZbin channel. Please take a look.

[@bob456fk6]

Re-coding a message before resending it doesn't seem reasonable. The operator at the source already had the characters written on his notepad, why not just send the same characters again via morse? For a message to be successfully decoded, Every character had to be received correctly. Interference was always a possibility. I imagine it was not uncommon to send the same coded message twice to make sure it was received exactly right. They should have had some sort of "check code" so the receiving operator could be sure he got the right code.

[@alexstevensen4292]

one thing I don't understand, the german text is punched on papertape. when the operator has to re-send the text he just has to run the tape again?

[@JC-jv5xw]

That's what i thought - The FSK radio signal could only be transmitted at the rate we heard by replaying a punched tape. Much more efficient use of the channel.

[@granskare]

In USA, we call valves = tubes.

[@v8pilot]

And the French call them lampes. Except for rectifiers which they call 'valves'.

[@lawrencebishton9071]

22 rpm why

[@lawrencebishton9071]

6inch lift on reflector Composit Mat tier ALLIMINIUM

[@Ogaitnas900]

NORMALIZE THE VOLUME! Please, you're going to hurt people's hearing! That burst close to the 23 minute mark is terrible. Anyway, amazing conference, thank you.

[@briancase6180]

The claim was made (here or in another talk, I can't remember) that the Colossus exhibited "parallel processing" because it could be processing two tapes at once. Actually, in this respect, it exhibited the technique of pipelining (the assembly line) because while one tape was being worked on, another tape cold be loaded and made ready. Pipelining has the effect of improving throughout like parallel (multiple) processing, but they are distinct implementation techniques. Colossus likely exhibited parallel processing is other ways. The bombe used parallel processing in the sense that it was seeking matches on more than one string at a time (lots of the rotors acting at once). I'm guessing the Colossus used a similar kind of parallelism, but I would need to investigate to be certain. Perhaps someone with definitive knowledge will make a comment.

[@ohgosh5892]

" it could be processing two tapes at once"; I think you may have been misled. My understanding is that the single tape of '5 bits plus Synch' was processed word by word. The synch stream was electronically read and used to clock colossus, so the tape could be run at any speed and the machine would follow. The tape reading was all electronic, light & photocells, and it is claimed that the tape reader was the fastest ever made. The 5-bit code was thus read on a word-by-word basis, and parallel processed, id est, word by word. These days we would call this a data bus with an 5-bit ALU, or something like that! The bedstead, the name for the tape handler, was huge, and there was, afaik, one per Colossus. There were several of them made, so messages could be parallel processed in that way.

[@briancase6180]

@@ohgosh5892 it was mentioned somewhere that a single Colossus could have two tape readers attached and functioning, but that this was not, in the end, the typical configuration. If I recall correctly, it was said that many people, operators for example, never knew that a two-tape configuration was possible because they had never seen one in practice or in theory.

[@ohgosh5892]

@@briancase6180 "it was mentioned somewhere that a single Colossus could have two tape readers " do you know what the second reader would be used for? It seems unlikely that there is sufficient processing power to analyse two input streams for a preponderance of XOR matches, so what would the second tape do?

[@briancase6180]

@@ohgosh5892 as I said, the second one was able to be set up and be ready to read while the first was reading. Then, there was no delay to start the serving tape. Pipelining.

[@ohgosh5892]

@@briancase6180 You actually said "because it could be processing two tapes at once.", but now you are saying that it couldn't, and they were delivered serially. You have changed "parallel processing" to "pipelining". I do not believe this, sorry, but if you can evidence this, I shall be happy to change my mind.

[@briancase6180]

Actually, the rotors moving together made this not a great (secure) encypering machine. The designers made several mistakes; some were subsequently ameliorated but the fundamental flaws were always there.

[@stanrogers5613]

The rotors moving together was absolutely fine. The problem was that they didn't always move together - the μ-wheels were the Lorenz machine's Achilles heel. It was the "stutter" that made the cardinality of the other wheels possible to discover. Apart from that (and the initial infrequent resetting of the χ- and ψ-wheels), it was only depths and partial depths that provided a way in - and that was mostly operator problems before the QEP system became standard.

[@briancase6180]

Colossus was fabricated with American threaded nuts and bolts because those were the only ones available in war time! The metric-threatened hardware was reserved for military (planes and guns, e.g.) hardware that engaged directly with the physical enemy. ☺️

[@ohgosh5892]

"Colossus was fabricated with American threaded nuts and bolts" No it wasn't. It was fabricated at Dollis Hill in London. The lecturer was talking about the Lorentz machine, made in Berlin. If you have any evidence of British use of metric threads in WW2 in the UK, I'd be interested to see it, as I've never heard of such a thing. As I'm sure you are aware, the original thread was the Whitworth thread, as invented in Manchester. Here is some more information on British threads: www.historywebsite.co.uk/articles/DarlastonIE/nutsbolts.htm

[@davidfaraday7963]

@@ohgosh5892 When I worked at Dollis Hill we used BA (British Association) series nuts and bolts. I never needed anything larger than 0BA but had I needed to do so I'm sure they would have been Whitworth. I can't, of course, say what was used on Colossus but I'd be very surprised if US threads were used as Colossus was constructed using mostly standard parts from the Dollis Hill stores which would not have included US threaded nuts and bolts. When the PO Research Dept moved to Martlesham Heath around 1970 the opportunity was taken to change to ISOmetric threaded nuts and bolts.

[@andrewmawson6897]

@@davidfaraday7963 I was working for Ferranti at that time, often visiting Dollis Hill where s team were working on Optical Post Code recognition based on a Ferranti Argus 500. iirc Tony Bass was one of the leading lights. Whole project moved to Martlesham - as I recall in those early days the canteen was still the original Officers Mess from WW2. BTW don't forget that BA threads were 'sort of' metric

[@davidfaraday7963]

@@andrewmawson6897 My brother also worked at Dollis Hill in the late 1960s and was involved in the OCR project, so you may have met him. ("Faraday" is a pseudonym for the purposes of YT, my surname is actually Looser). I remember seeing Doc Coomes around the site, his eccentricities made him instantly recognisable, though I never actually got to speak with him. My Dad had also worked at Dollis Hill during the war, and knew Flowers, though clearly had no notion as to what Flowers was working on. As my Dad said after the existence of Colossus was finally made public, at that time you simply never discussed your work outside of your team.

[@andrewmawson6897]

@@davidfaraday7963 My work with Ferranti took me to some interesting places that only in later year did I appreciate the significance of. So as mentioned Dollis Hill, Martlesham, Bletchley Park, GCHQ, FCO - trips to Cheltenham made me hanker after an RA17 which I still have! And indeed it was the chance acquisition of an HRO that lead me to decoding RTTY that probably secured the Ferranti job - after the usual 'what are your interests' question I probably spouted on for hours about the technicalities of decoding RTTY / Murray code - I think the only technical question I was asked was 'what is a ring counter'

[@allanegleston4931]

too bad they didnt have transistors and ics . this machine is the gr,gr,gr,gr eat grandad of my mac that im typing to u on this video.