BIND 9 AAAA record problems

[Todd T. Fries]

If the application choses its own error prone mechanism that does not
fallback to v4 when v6 fails then there are issues the application users
should loudly protest.

Anything that uses getaddrinfo should program a routine similar to
ssh_connect() in openssh .. namely a 'for()' loop that loops through
addresses for the host and trying each one in sequence until the
connection succeeds...

Anything else is just asking for trouble in a dual stack world.

On Tue, 2004-06-29 at 16:37 +0900, JINMEI Tatuya / 神明達哉 wrote:
> >>>>> On Mon, 28 Jun 2004 14:55:32 +0000, 
> >>>>> "Todd T. Fries" <todd at fries.net> said:
> 
> > Why is it a bad thing to start out with AAAA?  Proper getaddrinfo()
> > implementations should query first an AAAA then an A record for names,
> > so that applications can then connect to whatever is appropriate.
> 
> > It is all part of a `seamless' transition to v6 that has started, albiet
> > slowly.
> 
> > For whatever it is worth, the libc in OpenBSD behaves this way, and I
> > quite am pleased that it does.
> 
> > It's not a bug, it's a feature ;-)
> 
> In general and in theory, I agree.  But we should also note some
> practical issues in the real world.  To name a few:
> 
> - there are some misbehaving authoritative servers against AAAA
>   queries.  See draft-ietf-dnsop-bad-dns-res-02.txt for more details.
> 
> - I've noticed some applications do not implement the proper fallback
>   mechanisms from IPv6 to IPv4.  Some do not fall back to A queries
>   when AAAA queries fail (with "NONAME").  Some do not try to connect
>   the IPv4 addresses when the attempt to connect to the IPv6 addresses
>   fails.  I see mozilla/firefox show this behavior with its internal
>   resolver routines.
> 
> 					JINMEI, Tatuya
> 					Communication Platform Lab.
> 					Corporate R&D Center, Toshiba Corp.
> 					jinmei at isl.rdc.toshiba.co.jp