Re: [BEHAVE] using IPv4-mapped DNS/DNS64 resolvers for dual-stack/IPv6-only hosts

Hi Dan,

IMO, whether normal DNS or DNS64 should be used depends
on what applications residing in the host would be used.

Assume there are 2 applications installed on a DS host.
one is DS suitable application (e.g., firefox browser),
the other is IPv6-only application. When the DS suitable
application is used, normal DNS is prefered. When the
IPv6-only application is used, DNS64 is prefered.

In that regard, detecting the type of the hosts
make little sense to me. applications used rather than
the host they reside in matters to this issue.

Thanks,
washam

----- Original Message -----
From: Dan Wing <dwing@cisco.com>
Date: Tuesday, March 9, 2010 11:21 am
Subject: Re: [BEHAVE] using IPv4-mapped DNS/DNS64 resolvers for dual-stack/IPv6-only hosts
To: 'Zhen Cao' <zehn.cao@gmail.com>
Cc: behave@ietf.org

>  
>  
>  > -----Original Message-----
>  > From: Zhen Cao [mailto:zehn.cao@gmail.com] 
>  > Sent: Monday, March 08, 2010 7:13 PM
>  > To: Dan Wing
>  > Cc: behave@ietf.org
>  > Subject: Re: [BEHAVE] using IPv4-mapped DNS/DNS64 resolvers 
>  > for dual-stack/IPv6-only hosts
>  > 
>  > Late catch-up with one comment on the approach:
>  > 
>  > if a dual stack host sends a query to first DNS (::ffff:192.0.2.1) 
> but
>  > fails to get any resultsr, it will failover to the second DNS 
>  
>  Right.
>  
>  > and there will be the same problem that traffic will be translated 
> 
>  > via NAT64?
>  
>  Yes.
>  
>  So that means the first DNS server better be working.  
>  
>  I expect the industry can accomplish that.  _One_ way to accomplish 
> that is to
>  operate both servers on the same physical server, listening on two different
>  IP addresses, and responding differently based on the incoming packet.
>  
>  
>  So far, based on email discussion, this is the only viable idea that 
> does not
>  require modifying the host.  (I consider the DHCPv6 manipulations in 
> my draft
>  to be non-viable, even though they do not require modifying the host).
>  
>  -d
>  
>  
>  > Thanks,
>  > Zhen
>  > 
>  > On Sat, Feb 13, 2010 at 4:00 AM, Dan Wing <dwing@cisco.com> wrote:
>  > > A few weeks ago there was an active thread on BEHAVE and 
>  > 3Gv6@ietf.org,
>  > > worrying about how a network containing a mix of dual-stack 
>  > hosts and
>  > > IPv6-only hosts would work.  Ideally, to prevent 
>  > unnecessary address family
>  > > translation, the dual-stack hosts should use a 'normal' DNS 
>  > resolver and the
>  > > IPv6-only hosts should use a DNS64 resolver.  However, it 
>  > is difficult to
>  > > detect if a host is dual-stack or is IPv6-only.
>  > >
>  > > So, I wrote an Internet Draft describing details of several 
>  > approaches that
>  > > had been discussed on the mailing list, and attempted to 
>  > provide a list of
>  > > advantages/disadvantages to each.
>  > >
>  > >
>  > > While writing them down, I came up with another approach 
>  > which uses an
>  > > IPv4-mapped address in the list of IPv6 DNS servers.  An 
>  > IPv6-only host cannot
>  > > use that address, so it would use the next DNS server on the 
> list; a
>  > > dual-stack host can use such a DNS server.  Details are in:
>  > >
>  > > 
>  > http://tools.ietf.org/html/draft-wing-behave-dns64-config-02#s
>  > ection-3.1
>  > >
>  > > Please comment on this approach.  For whatever it's worth, 
>  > the idea seems to
>  > > work.
>  > >
>  > > -d
>  > >
>  > > _______________________________________________
>  > > Behave mailing list
>  > > Behave@ietf.org
>  > > https://www.ietf.org/mailman/listinfo/behave
>  > >
>  
>  _______________________________________________
>  Behave mailing list
>  Behave@ietf.org
>  https://www.ietf.org/mailman/listinfo/behave
>  