I've come to the conclusion that "character" is worth abandoning as a coherent concept. Words and bytes are much more useful and easier to define and aren't dependent on which runtime you're using, and the number of times it's worth slicing a word into characters is actually pretty low. One exception to this might be chinese (and I'm sure others) where differentiating words is not a straightforward task of splitting by whitespace, but they also have much more straightforward glyph rendering bypassing most of unicode's nastiness. Random access strings in chinese are actually super straightforward once you abandon run length encoding to heavy users of ascii and emoji.