Oops, I stand corrected. Encoded-variable as currently defined does
indeed suffer from the limited-number-of-headers problem. However, it
could be otherwise -- in particular, we could probabaly allow it in any
quoted-phrase or comment, thus making it more open-ended.