Good Cryptome work
Now, we remember Rosser from reading on Turing’s work (at Princeton) in the 1936 period. Turing’s Princeton connections are discussed at https://yorkporc.wordpress.com/2012/10/09/turing-scrypto-formation-while-at-princeton/. Modern relations between PRinceton and VeriSign founders is also quite interesting, as was their and Whit Diffie’s personal access to NSAs (rather dreary, “re-purposed architecture”) Finksberg bunker.
A brief précis of Rosser’ life work and direction speaks volumes (for why he would be selected as the probably ideal initial director for an NSA “academic” research group tasked by NSA’s own applied research staff):
In terms of Rosser’s background (in logic, computability, etc), the Professor is a part of the founding family.
Armed with a CDC computer, under Rosser early IDA folks evidently spent lots of time evidently building custom tables using numeric analysis techniques of the day (for fast computation of certain problems studied specifically to under the complexity of both problem and classes of solution):
Now, what we must NOT forget is that such academics – censored. These are the same kind of tables used by Rosser when helping build guidance systems for ICBMs, etc. But within 10 years, folks have gone from the research-capability CDC computer doing fast table generation to using a Cray-1 – for large scale data mining.
So what would an academic-focused research group be doing with that class of machine (given its so precious in terms of knowhow, manufacturing techniques and capability and being stored at Princeton in a non-physically secure facility)?
Now quite missing in the overall (pre) NSA history (i.e. the WWII era) is the thinking on machine-based attacks, as in:
While its clear US is doing fine with machinery for making ciphers and testing their outputs using “tabulators”, it doesn’t seem to have gone beyond the photoelectric effect for cryptanalysis.
While it is pretty clear that the pre-NSA ASA is tied tightly in with Bell, MIT. Lincoln and co, the WWII machine group’s “philosophy” of design is not well documented. One can see the RAMs (e.g. the 5205 built for Tunny) as first attempts to generalize what were entirely specialized machines doing coincidence counting (replacing hand methods, involving 5 million folks…doing hand counting) for particular emulations of cipher machines under attack. But, how folks thought about the genralization is not clear. Were they being advised by the Princeton group (who had all the same concept set re machine as did Turing, for obvious reasons)?