SUBJECT: Fish Notes Report #F 7
TO : CO, SSA, War Dept. 23 March 1944
1. As Capt. Seaman has doubless explained Mr Newman's section
handles the machine work on this problem and Maj. Tester's the hand
work. Since I am visiting both groups my reports on the techniques
observed cannct follow the order employed in the actual work of
solution. It might be well, therefore, to indicate the usual procedure
applicable to the Bream traffic (or any other traffic using the P5 limitation).
2 At the beginning or the month the X patterns must first be
recovered by statistical methods. This is handled by Mr. Newman's
section, The methods used are known at A. H. and have been reported
on by Capt. Seaman. I will report further at a later date. Next the
X patterns have to be set on a message which is one of a pair in depth.
The setting is also done by Mr. Newman's section. Some of the mathe-
matics involved win be recapitulated in my next Fish report. In this
operation both messages can obviously be used if one is not long enough.
After this has been done the de-chi is sent to Maj. Tester's group to
find the Psi patterns. This is not simple because the so-called depths
are not true depths by reason of the P5 limitation. The Psi patterns
emerge with plain text in the two messages and also lead to MT. This
must then be resolved into Mu61 and Mu37l. This last step is described in
paragraph 3. Later in the month when both X and Psi patterns are known,
Mr. Newman's people take all messages of sufficiant length, set the X
patterns wherever possible, and send the de-chies to Maj. Tester.
Depths are now not required because agreement with known pattorns furn-
ishes the check by which plain text assumptions are verified. Both
the original recovery of the Psi patterns and the setting of known
patterns will be described in due course. The setting of the Psi pat-
terns again yieIds MT which must be resolved as heretofore. When the
day's motor patterns have already been found, this last is a simple
problem because it is only a matter of finding the starting point.
One of the real difficulties of this entire problern arises from the
fact that it never reduces to mere decoding. Each message requires
independent cryptanalysis unless it is on the same day as one already
solved, on the same circuit, in the same direction. and with the same
QEP number - in other words, unless it would have been in depth with
a solved message were it not for the P5 limitation. Identical QEP
numbers in opposite directions mean nothing because different tables
of settings are used at the two ends of the circuit. The tables are
also different for each day.
3. The method used to determine Mu61 and Mu37 from MT is funda-
mentally the same as the old method with which .A. H. is familiar.
However, since MT and MB are not identical the probleni is now more
difficult and ordinarily requires at least 400 - 500 letters of text.
MB is first derived from MT by carrying down all values where X2 one back +
P5 two back = x. The other positions must be left blank because there is no
way of knowing whether the cross in MT derives from a cross in MB or from
the limitation. The broken MB pattern is written on a width of 61 and