Re: Replacement for C-x 8 based on input methods

[Richard Stallman]

    My current plan is to take latin-1-prefix and latin-1-postfix and the
    current C-x 8 keymap and to construct the union of all the sequences
    found there.

Doing it by hand would be fine if we wanted just one of these.
But we want them for each latin-N alphabet, and maybe more as well.

What we need is a mechanism for defining an input method whose
contents will be computed the first time it is activated.  With that
mechanism, we can define latin-1-bothfix to be computed as the union
of latin-1-prefix and latin-1-postfix.  And likewise for
latin-N-bothfix.