Steve Downey:
> We're probably going to want to decode utf-16 and 32 (convert to scalar
> value / code point) for a long while, but encoding to them should be
> unusual. At least that's the ietf and whatwg recommendation for new
> protocols and formats.
I've just checked and full UTF-16 encoding and decoding code in my
Unicode library is 104 lines + 28 lines for strong code unit type. So
the maintenance burden is very small.