Punycode (the way Unicode maps to DNS names) is pretty clever. Characters for a given language are close together, so Punycode optimizes for encoding characters close together. https://en.wikipedia.org/wiki/Punycode

Here's an example from the RFC: https://tools.ietf.org/html/rfc3492#section-7.1

    Unicode:      他们为什么不说中文 
    DNS:          xn--ihqwcrb4cv8a8dqg056pqjye
    utf8 base64:  5LuW5Lus5Li65LuA5LmI5LiN6K+05Lit5paH

Let's see if HN will link these:

http://.ws -> http://xn--ls8h.ws/

edit: Nope, it strips (!) emoji from comments.