Mark Snesrud and Bob Mayo explained how they solved the first Adobe semaphore puzzle twenty years ago. Near as I can make out from their paper, they assumed that it wasn’t the semaphore positions that were the code, but the changes from one position to the next. Each semaphore has four possible positions (at the time, anyway), but there are/were seventeen possibilities for each change for each semaphore – the semaphore can turn clockwise or counterclockwise, 45 degrees, or 90 degrees, or 135 degrees, etc. Like the Adobe website says, there are 256 possible semaphore displays, but if it’s the changes that carry the code, then each change has 17-to-the-fourth possibilities. But seems like Mayo and Snesrud immediately decided it was the changes that carried the code. How did they do that?
Simpler question: how long did the semaphore take to display the whole Pynchon novel?
They saw a pattern… they made figure 8 … it didnt really matter if they did this as the absolute or the delta .. the absolute would also remain fairly flat.. they would have seen that even if they only plotted absolute values… but in figure 8, they found a header…
Then in the header, they saw the pattern stayed the same for each header, but one bit changed a bit, it was a counter , so that shows the encoding of values … then they could know how what 0 to 9 encoded as, when every 10th step the 2nd position changed… the first position was therefore 0 to 9 …
They saw a pattern… they made figure 8 … it didnt really matter if they did this as the absolute or the delta .. the absolute would also remain fairly flat.. they would have seen that even if they only plotted absolute values… but in figure 8, they found a header…
Then in the header, they saw the pattern stayed the same for each header, but one bit changed a bit, it was a counter , so that shows the encoding of values … then they could know how what 0 to 9 encoded as, when every 10th step the 2nd position changed… the first position was therefore 0 to 9 … but they were ascii , so the whole header was decoded as ascii revealing the Ulysses text words as well as a counter.
Using this encoding as a precedent, they looked at the payload section and found the keyword would lock in certain areas of 0-255 in use, the keyword and AM radio code was setting the offset into the 255 to be removed… they just had to find similar keywords and AM codes, then evaluate what each step did. eg if everything was the same but FORGIVE changed to FORGAVE, then they could evaluate what keyword position 5 did to the offset into 0-255…they could know the precise offset since the decoded values had to produce the english frequency of usage of A to Z…