Re: [idn] UTC feedback

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

well, more generally, if

in language A it is important to be able to register two different 
IDNs that differ only in their use of accent marks

in language B it is important that one NOT be able to register two
different IDNs that differ only in their use of accent marks

then the only way I see to handle both cases is for 
each zone to enforce the rules needed for that particular zone.
this could be implemented in the servers themselves (they could 
refuse to accept a zone file that didn't implement the local policy) 
or in some other process that handles registration in that zone.
(note, however, that it's not just a problem for the TLD registries,
which argues for doing this in the server)
but it doesn't have to affect the IDN protocol itself.

on the other hand, if language B' has the characteristic of language
B, but in B' it's also important that one be able to type in a name 
with or without accents and reliably get to the same service, then 
there needs to be some mechanism that assures that this will happen.  
again, this could be implemented in the registration process for that 
zone (which, when given any name which contained an accent, 
automatically registered the version without any accents as a CNAME 
to the version with the accents) or it could be implemented as an option 
within the master server itself, (type an accented name in the zone file 
and it automatically infers a non-accented record with a CNAME pointing 
to the accented one) small sites would probably appreciate the ability 
to do this.

one wrinkle is that whatever way you came up with to solve this problem
has to work with DNSSEC.  so you may end up signing additional records.

can we satisfy ourselves that we need only deal with 

a) the word spelled with all accents, and
b) the word spelled without any accents

for all languages, or do we also need to deal with every possible 
combination of a letter being accented or not?

Keith

all of the above presumes that we're using a modified version of DNS
rather than some new protocol.  if we do it with a new protocol then
we could investigate alternate approaches.