Re: [Bundled-domain-names] DNS resolution for bundled names

["John Levine" <[e-mail]>]

>We hope to gather some interests for this issue to see whether we can do some standardlization work for it  in  IETF.
>Out initial plan is to have a bar bof in Berlin IETF meeting.

Thanks for taking this on.  It seems to me that the different bundling
scenarios can be addressed in different ways, and we might want to deal
with the easy ones first.

The easiest is probably parallel TLDs like .ONG/.NGO.  There's already
DNAMEs at the top level for the traditional and simplified versions of
Taiwan, and as far as I can tell, the DNS aspects work pretty well.
(As opposed to whether mail, web, and other servers do what you'd want
when the DNAME sends traffic their way.)

The .CAT domain bundles 2LD names, in their case an accented version
and ASCII version of the name.  They use a DNAME for the accented
version that points at the unaccented version, which works really
badly.  Partly that's because the DNAME doesn't redirect itself,
partly I've found from spot checking that even the 3LD names that the
DNAME does redirect rarely work in practice, with almost none of web
servers I checked handling them.

So one part of the work could be looking at BNAME or something like it
that could make parallel 2LD names work better, both within the same
TLD and within arbitrary places in the tree.

A separate issue is how one might arrange to have servers for web,
mail, and so forth automatically configure themselves to do something
reasonable when a BNAME or DNAME points at them.  Roughly speaking, when
they get a request for a domain they don't think they handle, they could
do a DNS lookup on the name and if there's a BNAME or DNAME redirect,
treat it as equivalent as the redirect target.  I realize that in some
contexts this has all sorts of security issues but I think it'd be
possible to carve out a useful subset that's not too dangerous and
still useful.

R's,
John