Should UTF-16 be considered harmful?

The problems not so much UTF-8 vs UTF-16 as the fact that there is so much software out there that can't handle unicode codepoints outside the BMP.

For example: https://github.com/rails/rails/issues/3727

Another example is MySQL. utf8 won't work; you've got to use utf8mb4 instead. (OT, but won't this be a problem again if and when we get unicode characters that are 5 bytes in UTF-8?)

I think the most amusing case of UTF-16 considered harmful would possibly be the choice, in WPF/Xaml, to use UTF-16 code point offsets to reference positions in XML strings -- while supporting UTF-8 XML encoding. It can create the situation where, even if the input and output are both UTF-8, you'll nonetheless need to transcode into then out of UTF-16 in order to perform the proper string splitting.

Fun.