10 minutes to try all settings and record those on which the count exceeds
a predetermined criterion. Here again / is the usual test, not the only
one. Plans are under way for runs which will count / against motor dots
together with some other letter (not 8 however) against motor crosses.
Self-match.or "slide" causes some difficulty in this operation. The P5
limitation does not eliminate this method although it increases the length
of message required. In general, however, if the message is long enough
to set the X patterns, there is a good chance that it is long enough, even
with the P limitation, to set the motor.
10. After the motor is set the Psi patterns can also be set. with
Colossus. They are extended by the motor and 2 Psi wheels at a time are
tried at all possible starting positions. Ideally, Psi1 and Psi2 should be
run against the de-chied message because P1 = P2 is a very strong plain
text characteristic. A number of considerations make this ideal
procedure impracticable (and sometimes impossible). It is impracticable
with present equipment because it takes too long to set up new patterns
on the machine. Two sets of patterns can be setup on the existing
Colossus and it is possible to switch either one into use. However, at
the moment, both Bream and Jellyfish DeltaX patterns are set up. Therefore
the procedure would be to make a run which would require that
DelatP1=DeltaP2. This is not as sharp but is still good enough. When the
new Colossus arrives (about June first) this practical difficulty will be
remedied. In fact Colossus #2 could do the whole job itself because it will
hold 5 alternative sets of patterns. The impossibility of proceeding in the
manner indicated arises when the P5 limitation is present. It is Then
necessary to set Psi5 first and the best run is P4 = P5. For the reason
stated above DeltaP4 = DeltaP5 would be the actual run. After Psi4 and Psi5
are set short runs are made on other Psi wheels with Psi5. Different plain text
characterIstics must, of course, be looked for on the various types of
runs. All of'the material set forth in this paragraph is, to some extent,
theoretical because Colossus has never been used to set the Psi wheels.
Some of the older machines have been used but the fact that it has all been
thought out and developed furnishes another instance of the advance
planning which seems to account for a great deal of the cryptanalytic
success which has been achieved here. Because of the possibility of
corruption (and its devastating effect when the P5 limitation is employed)
this type of run is limited to 800 letters at a time. The correct solution
is found through matching up the high scores. Experience shows that
about 1 letter in 800 is garbled on the fifth impulse.
11. Maj. Tester's section has derived the X and Psi patterns for
March Stickleback (X2 limitation but not P5) but has had no time to devote
to efforts to read the traffic.