I took part in the FJPEM research project of WIDE Project in the fall of 1993. At that time, Mine had implemented mhpem to make it convenient to use FJPEM with mh-e. Inspired by mhpem, I spend many time to enhance mhpem as escape from my master thesis in winter 1994. :)
mhpem was a lovely program to decrypt cipher messages automatically, however, we met some problems. The biggest one was that mh-e was not flexible enough to enhance some features. I really wanted to cache decrypted message to display it quickly for the second time but it was quite hard to enhance mh-e to do so. When a new version of mh-e was released and it appeared that mhpem didn't work with it, I decided to kick it out.
I want to cache decrypted PEM. Also MIME. Why isn't there a good interface to handle MIME with easy operations? Tell me the reason not to be able to cite multiple messages to one replying message. It's ridiculous that we cannot enjoy cheerful marks. I'd hate to refile messages with difficult operations. All in all, it appeared impossible for me to implement what I want by modifying mh-e.
Auto decryption in mhpem, asynchronous scan in mhasync, dynamic window setup in GNUS, cheerful mark system for multiple messages in gnus-mark, refile feature in VM, message cache mechanism, and beautiful and flexible programming style.... A pieces of puzzle was gradually getting together in my mind. It was early in the spring of 1994.