message settings now known to find the corresponding trigrams (Verfarhen kenngruppe).
(4) These trigrams gave partial equivalents for the bigrams sent in the indicator e.g. if the indicator was IJTV USYX (see example in Para. 27) and trigram had been found to be ARQ then IJ = ?A, TV = ?R, US = ?Q.
(5) Using the fact that the bigrams are reciprocal we can find the other half of the substitution e.g. having IJ = ?A, if we have also somewhere else LA = ?J then probably IJ = LA is correct. This gives rise to some alternative values but these can be cleared up by using the fact that the top trigram (Schluessel kenngruppe - YVT in my example) must come in a column in the K book corresponding to the key on which the message has been sent.
51. There are 9 tables and 3 days' traffic on a table was needed to clear it up completely. Until the tables were more or less cleared up (1) Banburismus could not be done (2) there was considerable delay in reading individual messages. The net effect of a bigram table change was therefore to dislocate our work on cribs fairly seriously for two to four weeks. Since we were entirely dependent on cribs during this period a change was much more serious in the early days when bombes were scarce than in later days when they were plentiful.
F Offizier Keys.
52. A certain number of messages which were either (1) very important or (2) personal, so that the officer in charge did not want the operator to know their contents, were encoded by the officer himself before they were encoded on the normal key. When therefore a daily key had been broken, a few messages were still unsolved as the "Offizier key" had still to be discovered. This key consisted of the following elements
(1) The same W.O. and Ringstellung as the main key
(2) An entirely independent stecker
(3) A set of 26 starting positions valid for a month and denoted by the letters of the alphabet A to Z.
53. For example, a message (when decoded on the main key) might begin with an address followed by "Offizier Sophie" and then go off into nonsense.