It will probably be a limitation of the underlying software stack, perhaps how repos are replicated. There are still users out there with OSes which do not do UTF-8 very well.

Also it can still be a problem guessing the correct encoding for user-provided data (some browsers are terrible at this).