RE: [idn] IDN character repertoire, nameprep

[Brian W. Spolarich]

| Is this an argument in favor of checking for prohibited names in client
| applications?  Today's client applications don't check for prohibited
| names (like a+b.com), and this hasn't led to any catastrophes.  Today
| canonicalization and prohibition are both performed on the server side.
| Nameprep proposes to perform both functions on the client side.  I'm
| suggesting that a less drastic change (performing canonicalization
| but not prohibition on the client side) may be sufficient, and even
| preferrable.

  As I recall BIND (which as I pointed out is an implementation of the
spec, not the spec itself), makes the checking for prohibited characters 
in labels optional, so this is not without precedent.

  -bws