The bytes "beyond JAPH" are not zeroes for I tried to take advantage of the remaining 56 degrees of freedom. (The lenght of the string must be a power of two for the periodicity effect to take place.)

The wrong assumption I had initially done was that due to the belonging of 'a'..'z' and 'A'..'Z' to a restricted part of the set of all 256 charachters (especially "in terms of individual bits"), somewhere along the orbit of the transformation, and for a fortunate choice of the remaining bits I would have got a particularly "simple" string, for a sensible definition of "simple", i.e. something that I could code with ease in a sufficiently small space.

So the further unpack()s where not expected at all in the beginning. Also, I had thought of an extensive search over all 2^56 remaining 7-bytes strings, 63 odd rotate values and 127 iterations of the transformation. On a second thought, considerations inherent to the limited duration of the universe , human life compared to it, and computing power available, convinced me to try a random search instead (wrt the 7-bytes strings).

I got the string I'm actually employing after running my brute force search program for a while. It is nice in that (i) it yields a short enough coded string, (ii) it doesn't need escaping of chars (in some cases the string contains both "'"s and '"'s), (iii) it needs a resonably high number of iterations to reach the JAPH string.

It may be worth mentioning that I've been running the search program in the background for a few weeks on a cluster of machines I have access to and it hasn't found a single coded string shorter than the current one, let alone a "better" one. In this connection it may also be worth mentioning that notwithstanding the fact that it had been launched with nice 18, that it caught correctly signals and that it carefully avoided having more than one instance running on each machine, it caught the attention of a few idio^Wother users who complained about it, so that I had to stop it definitely.