Talk:Two-square cipher

A couple notes on conventions
From what I have seen, it is far more common to merge i and j in making the squares, rather than remove q from the alphabet. At least, the references I have seen all do this. I wonder if the examples here (and in related ciphers) should be changed accordingly.

Also, so far as I have seen, the different convention for same-row digraphs in horizontal two-square (reversing the digraph rather than preserving it) is universal, so I've updated the page to reflect this. I figure this had to be done, as otherwise the text in the cryptanalysis section made no sense. Mahousu (talk) 02:22, 29 May 2008 (UTC)


 * You should change it -- but for maximum historical accuracy it would be best if we could find Delastelle's original paper and see what he did. Of course which letter to drop is largely arbitrary, but it should be one which minimises the risk of confusion in the plaintext language in use. Whereas Playfair was largely used by the British and Commonwealth, Two-square is French, and in French the letter Q is not a particularly rare letter, being more common than 10 other letters including "H", "F", "W" and "B." But the difference between "I" and "J" is probably more important in French too, because the sounds of these two letters are much more different than in English. In French the most logical letter to drop is "K" but I don't know if that is what Delastelle actually chose. -- 203.20.101.203 (talk) 08:07, 13 July 2009 (UTC)

Two-Box Cipher
Several (all?) of the references in this article claim that the specific cipher actually used used in late 1941 (apparently officially called the "Doppelkastenschlüssel" - Two-Box Cipher) not only has 2 squares, but also "double-encrypts" each pair of letters in 2 steps: Step 1 involves finding the first plaintext letter in the *left* square and then finding (depending on the second plaintext letter) the first intermediate letter in the *right* square. Step 2 involves finding the first intermediate letter in the *left* square, then finding (depending on the second intermediate letter) the first ciphertext letter in the *right* square. (This kind of "double encryption" wouldn't work with a single-square Playfair cipher, much like 2ROT13). Also, the pair of letters picked from the plaintext are adjacent *vertically* (2 lines of text), which is why Noel Currer-Briggs spends a lot of time discussing words that are so long that the first letter of the word becomes paired with a letter near the end of the same word -- most words are so short that each letter in them is paired with a letter from *some other* word in the message.

Currently the article describes a simplified "single-encrypted" cipher skipping Step 2, and that picks pairs *horizontally* -- most pairs are from *the same* word.

Is there a specific name for the simplified "single-encrypted" cipher currently described in the article, that distinguishes it from the "Doppelkastenschlüssel"? Was that "single-encrypted" cipher actually used historically as anything other than an educational "toy cipher" in preparation for learning the more complicated cipher? --DavidCary (talk) 18:11, 26 March 2020 (UTC)