Q3: What characters should be allowed in a revised IDNA2008 specification?

[Gervase Markham]

On 01/04/09 14:42, John C Klensin wrote:
> This is, of course, the other (and, IMO, very strong) argument
> for banning non-language (i.e., other than an extended form of
> LDH) characters  in the protocol and specifically in lookup.  If
> we specify the characters that are rejected on lookup and do a
> competent job of it, then all conforming browsers (and other
> applications supporting looking up IDNA labels) will ban the
> same bogus labels.   If we don't, we leave it up to individual
> interpretation.  Such interpretations have all of the issues of
> inconsistency for users that have been raised for local mappings
> -- some strings work in some application contexts and not others.

And there is the high likelihood that client software will try and be 
generous and permissive, but yet as new problems are discovered have to 
get more and more restrictive. (That's certainly what has happened with 
our short list.) As this restriction process will be inevitably patchy 
and uncoordinated, previously-valid domain names will stop working by 
degrees as software gets updated. Ick.

Gerv