logues listed positions of the machines which would satisfy certain conditions which were implied by the indicator groups and from them the machine set-up for the day could be worked out.
On May 1st 1937 a new indicating system was introduced. The first 2 groups of the message were repeated at the end, thus showing clearly that they formed the indicator, but it was immediately apparent that throwing-on had been given up. This was a sad blow, but the Poles succeeded in breaking May 8th and they discovered the Grundstellung with the help of messages to and from a Torpedo boat with call-sign "AFA" which had not got the instructions for the new indicating system. On breaking May 8th the Poles discovered that it had the same wheel order as April 30th and the intervening days were soon broken. AFA's lack of instructions and the continuation of the wheel order are typical examples of good fortune such as we have often experienced and also of the German failure to appreciate that for a cypher innovation to be successful it must be absolutely complete. >
May 8th and the preceding days could not, of course, be broken with the catalogue of box shapes as the indicating system had changed and for them the Poles devised a new method which is of considerable interest. Their account of this system, written in stilted German, still exists and makes amusing reading for anyone who has dealt with machines. The process was fundamentally a form of cribbing, the earliest known form. On the basis of external evidence one message was assumed to be a continuation (a FORT) of another - apparently identification was easy and FORTS numerous in those days. The second message was then assumed to start FORT followed by the Time of Origin of the first message, repeated twice between Ys. German security must have been non existent in those days as these cribs appear to have been good and the Poles quote an example which, by its pronounceability, gave its name to the method of attack they had evolved - FORTYWEEPYYWEEPY = continuation of message 2330, numerals at this time being read off the top row of the key board and inserted between Ys. I will not attempt to explain the details of the method: it involved a series of assumptions to the effect that certain pairs of letters were both self-steckered and, if the assumption was correct, the method worked. At this time, it will be remembered 14 of the 26 letters were still unsteckered so that the assumption was not a very rash one and also the number of wheel orders was very small.