In the era of Unicode, this battle is pretty much lost. Several different code point sequences can produce the glyph 'ä', and user input can contain any of these. You need to normalize anyway.