Talk:KL-7

From Wikipedia, the free encyclopedia

MILHIST This article is within the scope of the Military history WikiProject. If you would like to participate, please visit the project page, where you can join the project and see lists of open tasks and regional and topical task forces. To use this banner, please see the full instructions.
Stub This article has been rated as Stub-Class on the quality scale.

I took out the link to Walker spy ring beause the story is told in the John Anthony Walker link. I don't think a second article would is needed, maybe redirect. --agr 15:41, 27 Sep 2004 (UTC)

Sure; I've added the redirect too. — Matt 15:45, 27 Sep 2004 (UTC)

[edit] Two rings or one?

Currently we have this description:

Each rotor consisted of a double ring of contacts. To establish a new encryption setting, operators would select an inner ring with 10 contacts and place it in the 26-contact outer ring at the correct offset as specified in the key list. This process would be repeated eight times until all rotor positions were filled. The outer ring permuted the message text; the inner ring controlled the stepping of the rotors in a pseudorandom fashion, a design principle that had proved successful with SIGABA.

Do we have a source for this information? I'm skeptical about having two concentric rings — the photo on [1] shows a single ring of 36 contacts. — Matt 12:42, 13 Oct 2004 (UTC)

I think you are correct. The photo does show 36 contacts on a single rotor, thought the caption says 26. One source talked about an inner and outer ring, but I now think that must refer to the plastic indexing ring. There are two sets of contacts on the basket (26 and 10) and the sources say that the 10 went to the stepping solonoids. I'm not sure how this works. One has to insure that a signal on the 26 pin input goes to a signal on the 26 pin output. There is also the question of how figs and ltrs shift was done. I've seen other sources that say SIGABA encoded space to "Z" and the KL-7 may have used a similar arangement. Anyway, I've changed the text, --agr 13:37, 13 Oct 2004 (UTC)

Thanks. Sorry to be nitpicky, but I'm also unsure about

a 10-pin set was used to control the the stepping of the rotors in a pseudorandom fashion, a design principle that had proved successful with SIGABA.

I presume this was based on this quote: "Each side of the rotor basket has two groups of 26 contacts and 10 contacts. The 26 contacts make connection to the circuit board and then onto the keyboard. The 10 contacts however make contact only with the KLA Stepping unit where an internal wiring harness routed the signal path back to the 10 contacts on the other side. This was called the reentry circuits for a more secure system" — [2]

Other parts from the above source and other sources suggest that the stepping was controlled by the outer plastic rings, which contain a series of indentations. My personal interpretation is although the 10 contacts make contact with the Stepping Unit, they don't contribute to the stepping; instead, they just pass through wires across it back to the other side and back into the rotor assembly again; this then results in a 26-element permutation (perhaps "26-point"?) using 36 contact rotors (this idea is illustrated at the very bottom of [3]). — Matt 14:27, 13 Oct 2004 (UTC)

I've been thinking about that too. I agree that the feedback scheme makes sense (is that your page?) but I also suspect it is used to control stepping as well. Elsewhere it is pointed out that the steping is controlled by 7 electromagnets. It would make sense to use 7 of the 10 feedback lines to step the rotors as well. That would follow the SIGABA design. The plastic rings are involved, but they are what the solonoids engage. Let me sleep on it. --agr 02:29, 17 Oct 2004 (UTC)

I have some more ideas. There are only 29 characters in Baudot that would actually participate in cryptography: blank (no punches), DEL (all punches) and BEL would presumabley not be used. They encode no information. There are 7 cam-controlled switches in the KL7 rotor box. It may not be a coincidence that 29+7=36, the number of contacts on the rotors. Perhaps only 7 of the group of 10 contacts are fed back. That would create a permutation of 29 elements, just right for encoding Baudot. The 7 feedback lines could also be connected through the switches to the rotor advance solonoids, the switch on one rotor would presumably control a different rotor. I don't know if the plastic cams were all the same, but if they weren't, that would add another keying element. For the purposes of the article, however, it might be best to say we don't know. --agr 14:53, 18 Oct 2004 (UTC)

Yes, I think for the article we need to be conservative about what we express as fact, particularly when there's not much available in the way of sources. (The web link is indeed my page, by the way). I like your idea about the 29-character alphabet, but I'm a little unsure about the stepping being controlled by the feedback lines. This would make the stepping dependent on the plaintext/ciphertext. There's a practical problem: recovery from transmission errors is much more involved. It's safe to assume that there'd be occasional corrupted characters in transmission. In other rotor machines that we know about, including the SIGABA, the stepping is fixed regardless of the plaintext. If a ciphertext character is corrupted then only one character is corrupted in the decrypted message. However, for a machine that uses plaintext to control the stepping, then a transmission error is much more difficult to recover from. Moreover, there's also a theoretical issue that some plaintexts would produce weak stepping patterns. For these two reasons, I'd be quite surprised if the NSA had employed this type of mechanism in the KL-7. — Matt 15:30, 18 Oct 2004 (UTC)

I agree that making the cipher dependent on plaintext is not a good design, but there are two eyewitness reports that suggest single character garbles did ruin the decryption. But then the garbles could have been multi character in nature e.g. "WT" for a Morse code "J". My comments about Baudot above are wrong, by the way. All punches is DEL in ASCII, it's FIGS in Baudot. And BELL is a FIGS-shifted character, so its code is not ignored. My 29+7 idea no longer adds up. I've edited the text to reflect the uncertainty, which is probably where things should stand. I did add a sentence about the permutor board shown on the Canadian web site. I'm trying to puzzle it out in detail. It seems to be slid back and forth by the knob to the left of the keyboard.--agr 19:59, 18 Oct 2004 (UTC)

A recent post on the Yahoo "cryptocollectors" group provides some more anecdotal evidence on this topic:

Special Note: If an incorrect letter was typed during Encypher or Decypher the KL-7 did NOT lose Crypto Key synchronization; it only produced one incorrect letter!! The rotors would step once or remain stationary each cycle depending on their "Notch" cams regardless of which letter key was typed.

If this source is correct, then the KL-7 stepping was independent of the message. — Matt 12:24, 30 Nov 2004 (UTC)

I'd like to contribute to this topic, as I see a number of errors. Particularly the reference to "Midnight", because of the use of "crypto day" instead. I'd also like to expand on "system indicator" and "message indicator", and on the length of messages. I have a problem though, and have not been able to find a solution. I usually enter Wikipedia by a link which does a LogIn. Yet, every page I go to it shows I am not logged in. If I take the link to log in, and return to the page I was looking at, it no longer shows me logged in. Anyone have a pointer for me to where I can get help on resolving this? If anyone wants to corespond and have me give them the information for the resolution of the above described errors, you can email me at w4crypto@walterjgould.com -- just mention Wikipedia. Jay69.19.14.42 20:16, 29 January 2007 (UTC)

Hi Jay, do you have "cookies" enabled on your web browser? See also: Wikipedia:How_to_log_in#Log_in_problems. Until the problems are resolved, however, you are still very much able to edit the page without being logged in. Please do let us know what errors we have on this page. — Matt Crypto 23:41, 29 January 2007 (UTC)