Definitions limit on label length in UTF-8

[Shawn Steele]

FWIW: I think that UTF-8 should NOT be a limit for Punycode.  How an app (or OS) encodes a decoded Punycode string internally is up to them.  I doubt we'd express such limits in GB-18030 or EUC-JP?

The only case I can think of for a UTF-8 limit is in the event someone made a UTF-8 clean DNS in the future.  However an ASCII punycode label is clearly not the same thing, even if it represents a similar string.  

In practice, all imposing a UTF-8 length limit does is to break IDNA2003 further, make it harder to code, and a little more error-prone.

-Shawn