Re: [idn] Alternative Solutions

[Keith Moore <moore@cs.utk.edu>]

> What started this thread was the possibility that patents might prevent
> us from using an ACE. 

We have to use some sort of ASCII name if we expect any IDN to be usable 
in an existing application, because those applications can only deal 
with ASCII names.  

I suppose it doesn't absolutely have to be an algorithmic mapping 
between the ASCII name and the native IDN, but I really don't like 
the idea of doing the mapping on the server.  I see two problems 
with doing the mapping on the server (in addition to those that 
you've already identified):

1. it's difficult to do the mapping in a way that is guaranteed to
produce consistent results between applications using the native IDN
and applications using the ASCII form of the name.

(without this guarantee, when people upgrade to IDN-aware applications
lookups of the same names that they previously used will sometimes 
produce different results - and fail at different times - than before)

2. (sort of like #1) it's difficult to do the mapping in such a way
that the server responsible for ascii->native mapping is the same as 
that for native->ascii mapping.

(without this guarantee, we open up an opportunity for there to be
different authoritative servers - and thus conflicting authorities -
for names that should be equivalent)

however if a solution can satisfy those two constraints, it might 
be worth a try. 

--

one way to do this might be to define a new RR type that specified
that a single ASCII label and a single IDN label (both relative 
to the current zone) were equivalent.  for instance,

#####	IEQ	xyzzy

would specify to the DNS server that (native) label ##### was
equivalent to ASCII label xyzzy.

labels in IEQ records could contain suffixes (and these might
be needed for glue) but only if left and right side had the 
same suffix up to the least significant label.  Thus

#####.a.b	IEQ	xyzzy.a.b

would be legal, but

#####.a.b	IEQ	xyzzy.c.d

would not.

if such a RR appeared in the example.com zone, the servers for
that zone would know to answer queries (regardless of query type)
for #####.example.com identically to queries for xyzzy.example.com.   
in either case one or more IEQ records could be returned as additional 
information, allowing the caller to discover the mapping between the 
native IDN and the ASCII equivalent.

IEQ RRs could either be explicitly specified on a per-zone 
basis or automatically generated by the master server.

Keith