What’s the problem?
One cryptological problem when trying to break this kind of code is that it’s impossible to discern letter boundaries of the plaintext. While it is true that each plaintext letter is always encoded by the same string of numbers, it’s impossible to say where one of those substrings ends and where the next one begins.
Furthermore, there are many degrees of freedom the encipherer has how to break down his plaintext letters. Take at look at the letters “m” and “n”, for example. In this example I’ve chosen to break them down in this manner:
This meant that the plaintext letter “m” was always enciphered as “2 5 5”, and “n” as “2 5” (Remember, the code “2” represents a short vertical stroke, and “5” looks like a walking cane with the handle pointing to the left; see previous page.)
But such a breakdown would be conceivable too:
We would have to introduce a new shape or stroke into our “cipher alphabet”, namely the mirror image of “5”. If we assign the code “7” to it, “m” becomes “7 7 2”, and “n” becomes “7 2”.
Or even such a thing is possible:
If the “arch” is introduced as code “8”, “m” becomes “8 5”, and “n” is simply “8”.
As you see, the possibilities are endless. (I’m not even going into the details of using CAPITAL or small letters, hand writings like batarde or print, etc.) And each choice affects the statistical properties of the ciphertext.
But what is even worse —
It’s well known that the “word” is a cryptological unit in the VM. That is to say, the ciphertext words are systematically composed, or, in other words, word boundaries in the ciphertext are not arbitrary: Why else would letter combinations like “qo” always appear word-initial, why would “dy” be always word-terminal? Subtle rules govern the composition of the ciphertext words (see my feeble attempts at shedding light into it). Each paragraph begins with a gallows character.
How can this be reconciled with the stroke theory?
Hardly at all. A VM word would be the equivalent to two or three plaintext letters. While it is conceivable that the plaintext was chopped up into two-letter pieces, it’s difficult to understand why those letter pairs would always start with special stroke sequences.
But then, I had an idea… see Mark II!