Pretty cool! I was curious about this, so I built a 50-line Python version to confirm my understanding. Feel free to check it out and use it (LGPL): http://bitbucket.org/carver/chaocipher/src/tip/chaocipher.py

It could use things like a configurable key and interactive enciphering command line, and I would be happy to pull such changes.

Edit updated link with improved comments about the rotation part of the cipher

Your code serves as a much better explanation than the linked paper in the article. And that Perl code o_O

Thanks! The paper was maybe a bit pedantic, but I appreciated his checkpoints and test cases. The perl was way overkill, though.

Why did you use state?

Cracked it

http://github.com/mavxg/chaocipher/blob/master/chaocipher.hs

(have not run, but from the description) :

Impressive work :) If you manage to get plaintext from just cyphertext, be sure to spread it around; I'd imagine a lot of people would be interested in it.

It doesn't quite do that yet. Given a section of known plaintext of about 50 characters it can find the original key used (or at least enough of the key to then turn the problem into a mono-alphabetic cipher).