> This does not mean that UTF8 may not be a good long term
> target, but before we can use UTF8 we need to have upgraded
> most of the net. It also means we need a intermediate
> target that can be used today so that those that care to
> upgrade will get the benefits now rather than in 10 years
> time.
So why would a fall-back scheme, such as I've sketched
before, not work? (*With* suitable security measures.)
One would then use only ASCII names (or IP number
'names') when 'talking' to instances of servers which didn't
(in some suitable way) declare that they can handle UTF-8
names. The UTF-8 name may be substituted with the ASCII
fallback along the way if not all steps can handle UTF-8
names. That would mean that sometimes the UTF-8 name is
kept, hopefully more and more often, as servers get upgraded,
sometimes not. But any fallback would be a name selected by
some person, not a more or less cryptic encoding, which WILL
sometimes FAIL to be decoded properly in principle AND in
practice forever. (A security point is also that any
cryptically encoded name, like in CIDNUC, that wasn't
decoded properly might not be considered secure by the
person that might use it, since the 'real' name is not
apparent.)
Kind regards
/kent k