“F9.1: a one-to-one letter substitution, then each hand transposition built since 0106 (compass routes, keyword shapes, keyed blanks, row reshuffles, French diagonal, EVA, Chiffre XIII, YAR, the plate layout, the poem code): exact”
F9.1. The list first: hand transpositions outside 0106's 1.86 million routes, each built since for F6, with its size: compass-direction readings and pairs (0159, 110,820), keyword-disrupted, triangle, interrupted and trapezoid columnars (0164, 1.40 million), keyed blank cells, in-row reshuffles and the French diagonal (0165, 4.08 million), EVA's double triangle (0166, 3,840), Chiffre XIII's shaped mask as a move (0166, 2.25 million), YAR's counting-off orders (0166, 24), paths over the copper's layout (0184, 73) and the poem code's double columnars (0190, 6,092). Not in this run: the 105.5 million mixed matrix triples (0158), 20.4 million keyword triples (0162) and 116 million Nihilist routes (0167), whose builders are compiled into their experiments' testers. Each route, forward and inverse, with 0106's test: a one-to-one substitution commutes with the move, so equal known letters must land on equal carved letters and different on different. Builders reused from the committed experiments; the test compiled; K4 against 10 shuffled copies.
Perhaps the letters were first swapped for other letters by a fixed code alphabet and then shuffled into a new order. Then wherever the messagethe message The 97 letters of English that Jim Sanborn wrote and enciphered. has the same letter, the copper must have the same letter after unshuffling. That simple rule was checked for 1.86 million shuffles before; here for every newer shuffle the project has built since, about eight million of them.
Words in the margin are defined on the glossary page.
1.1
The method at work
The carved letters go in at the top; the reading comes out at the bottom.
the idea
Swap the letters by a fixed code alphabet, then shuffle them: the order the Zodiac killer's 340 used. Then the same message letter always lands as the same carved letter, wherever the shuffle puts it.
the check
For each of ten million moves built since 0106, follow the 24 known letters to their carved places: the three Es must show one carved letter, the two As another, and so on, all different.
the bound
No move passes on K4 or on ten scrambled copies; a planted text does. The three largest compiled libraries follow in 0198.
2
Best candidate against the known letters
This experiment produced no candidate: the search ruled everything out before a full text was built. Its product is the bounded negative in section 7.
3
Walkthrough, by hand
Redo the method on a few letters with pencil and paper.
Take a route: it says where each message letter is carved.
Follow the three Es of EASTNORTHEAST and BERLINCLOCK to their carved places: all three must show the same carved letter.
The A, S, T, N, O, R, L, C must likewise each show one carved letter, all different from one another.
One mismatch kills the route.
About eight million routes, on K4 and on scrambled copies.
the worksheet: the first steps by hand, on squared paper
4
Worked example, generated
The same steps run by the code on the real letters.
the known letters and the carved letters they would have to match one to one, with no move at all:
place 22: E -> F
place 23: A -> L
place 24: S -> R
place 25: T -> V
place 26: N -> Q
place 27: O -> Q
place 28: R -> P
place 29: T -> R
place 30: H -> N
place 31: E -> G
place 32: A -> K
place 33: S -> S
place 34: T -> S
place 64: B -> N
place 65: E -> Y
place 66: R -> P
place 67: L -> V
place 68: I -> T
place 69: N -> T
place 70: C -> M
place 71: L -> Z
place 72: O -> F
place 73: C -> P
place 74: K -> K
5
Hypothesis, for a cryptanalyst
Family, key space, the constraint under test, and the clues it reads.
F9 (the Zodiac 340 construction: substitution first, transposition last; 0018 found it the only construction making the width-21 repeats), A9. Model: c = route(sigma(p)) with sigma any permutation of the 26 letters. 24 known letters, 13 different: 11 equalities and the inequalities between them, a strong test (0106 found no survivor in 1.86 million routes). Statistic: surviving routes, K4 against copies.
6
Prediction
Written before the run; the result decides it.
No route survives on K4 or on the copies, as in 0106. Zero candidates.
7
Method
Load each committed experiment's route builder.
Build the routes in parallel chunks, forward and inverse.
For each route and text: the carved letters at the 24 known letters' landing places must match the known letters one to one (compiled).
Count survivors per text; K4 against 10 shuffled copies; a planted text as a builder check.
8
Result and bounds
The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.