First of all, I do support a UDNS type proposal. And, I am working on an
I-D that enhances UDNS too.
Anyway...
> > - uses ACE for backward compatibility including supporting
> > IDNA clients.
>
> I'm not disinterested in compatibility with 37 (or 63) valued encodings,
> however, in its simplest form, a utf8-over-the-wire proposal need not
> make reference to any ACE. Can you reduce the problem to a simpler form?
>
It is not a complete solution if ACE is not referenced. What happens when
an application obtains DNS information (such as an A record) from the DNS in
UTF8 (probably via EDNS) and then try to contact a destination server (such
as http) that is non-IDN aware with the domain name? In this case, the
application must know what or how to use an appropriate ACE name to contact
the http server.
>
> > - uses EDNS to allow longer labels than 63 bytes.
>
> Extending the number of octets used to form labels is both very hard,
> and unnecessary to extend the set of valie values available to form
> labels. [I appreciate that the limit hurts, but length is disctinct
> from values allowed, and where possible, one problem at a time.]
>
Perhaps if we use UCS directly then the "count" can actually be the number
of Unicode code points and every language will have 63 "character" spaces to
use.
Edmon
