Another input method feature needed

[Stephen J. Turnbull]

Richard Stallman writes:

 > It would be nice to have a mode in which you can browse through all
 > the various character sets, move to a character you are interested in,
 > and ask how to input it.

A naive implementation won't cut it.  There are an *awful lot* of
characters, and browsing gets painful if you do it by Unicode blocks
unless you already know pretty much exactly what you're looking for.
Eg, "accented Latin" requires browsing through maybe 700 characters,
and they all start to look alike after the first two rows.  It's easy
to miss what you're looking for.

Besides information on Unicode blocks (which should be already
available), I would suggest using an NFD decompostion, and then allow
the user to select components (based characters and composing forms,
eg, turning the Angstrom symbol into an A and a circle).