José, danil and everybody,

  I was looking into the issue but got side-tracked.  I'm not ignoring
it but just taking long time.

  On XO, a composition character comes in as a separated character
code (the Unicode code point), and the semantics is that the
application composes the composition character to the character in
front of it.  The ParagraphEditor in the Etoys image is modified so
that as long as the resulting character has a valid code point, these
two characters are replaced with the resulting accented character.  In
the other words, it is a postfix manner.

  On typical Linux installation, however, the accent is prefixed.  The
keys are interpreted by scim or something but the Squeak VM doesn't
handle it.  There is a proposed patch for Japanese as well, and I'd
like to merge it and along the line, I'd like to fix the accented
character cases.

-- Yoshiki