Facebook over IPv6

George Bonser gbonser at seven.com
Mon Jun 13 17:54:01 CEST 2011



> -----Original Message-----
> From: Gert Doering [mailto:gert at space.net]
> Sent: Monday, June 13, 2011 2:07 AM
> To: George Bonser
> Cc: Ted Mittelstaedt; jkrejci at usinternet.com;
ipv6-ops at lists.cluenet.de
> Subject: Re: Facebook over IPv6
> 
> Hi,
> 
> On Sun, Jun 12, 2011 at 07:53:45PM -0700, George Bonser wrote:
> > For example, if I get a request from foo.com for an AAAA record for
> > www.some-domain.com that arrives over IPv4 when I have a v6 server
in
> > whois, I might treat that request differently than a request that
> > arrives via v6 for the same resource.  The reason is that if the
> request
> [..]
> 
> This idea keeps coming up again and again, and there still is no
> evidence
> that the protocol that a client's resolver uses to reach an
> authoritative
> DNS server has *ANY* relationship whatsoever to the protocol that the
> client itself might prefer to use to reach the server.

That depends. If the resolver being used is one of the public resolvers,
that is correct.  But in my particular case, the clients reaching my
network use their own network's resolvers as the DNS server is not
configurable by the user.  That applies in practice to many of the
general use situations, too.  Less than 1% of 1% of users (.0001%) are
going to manually configure a DNS server.  I also said I have *more*
confidence, not absolute confidence, which is why a v6 query will get
both a v4 and v6 response but a v4 request will get *only* a v4
response.

They could have perfectly working IPv6 and still not be able to reach me
via v6 due to routing "shadows" which is why the request to me came via
v4.  But if the request does reach me by v6 and the request is not from
one of the public resolvers, it is very likely that the user's provider
has IPv6 to their network and can reach me. At least there is some
chance that v4 could work.  If the request arrives via v4, I am not
going to assume they can reach me by v6 and will also assume that
handing the client a v6 resource is more likely to simply delay the
client in reaching me.



More information about the ipv6-ops mailing list