UTF-8 heavy languages like Japanese manage to pack in more meaning into 140 characters than English. Here is one farcical attempt to keep them honest: truncate the tweet after 140 bytes, not characters.

140_bytes_not_chars

javascript:(function(d){
var tbox, text, len = 140, ii = 0;

tbox = d.getElementById('tweet-box-global');
if (!tbox) return;
if (!(tbox = tbox.firstChild)) return;
if (!(text = tbox.innerHTML)) return;

while (true) {
  len -= encodeURI(text[ii]).replace(/%[A-F\d]{2}/g, 'U').length;
	if (len < 0) break;
	ii++;
}
tbox.innerHTML = text.slice(0, ii);
})(document);void(0);

Supposedly there is a lot of subtleties around UTF. This "encodeURI" approach is probably not perfect, but it seems to work with Japanese characters (which is the only language that I know besides English).