Re: [idn] I don't want 8-bit failures in 2011

["Eric A. Hall" <ehall@ehsco.com>]

I've read the material there, almost all of it anyway.

> And for a really long term solution (which may be totally out of the
> scope of this WG) due to the radical approach proposed,
> draft-klensin-dnsclass0e-00.txt. :-)

Doesn't work to extend the label space.

The question section has a fixed structure, so even with a new class the
Qname is still limited to 63 byte labels and 255 byte octets. You'd have
to do some ugly hacks to extend it (or use EDNS).

There are lots of other problems with new classes. CNAMEs across classes
probably won't work for example. All of the IN specific RRs (of which
there are many) would have to be reinvented/redefined. Etc.

Alternative class is creative but it brings too much baggage. EDNS is the
most effective proposal currently on the table for many reasons.

-- 
Eric A. Hall                                        http://www.ehsco.com/
Internet Core Protocols          http://www.oreilly.com/catalog/coreprot/