no IPv6 interfaces found

Jim Reid jim at rfc1035.com
Mon Aug 23 15:48:32 UTC 2004


>>>>> "Marten" == Marten Lehmann <lehmann at cnm.de> writes:

    Marten> If I'm telling BIND to bind to a certain
    Marten> interface and not to look for IPv6 interfaces, I'm
    Marten> expecting BIND to do this.

You are telling the name server to bind to a specific interface. But
you're not telling the name server it shouldn't look for IPv6
interfaces. There's no way of doing that, short of telling the server
is must not look for network interfaces after it starts. But even at
start-time the name server will still issue a SIOCIFCONF ioctl() and
get info about all the network interfaces on the host. Putting a
clause like:
       "listen-on-v6 { none; };"
in named.conf only tells the name server not to listen on IPv6
interfaces. It does not say anything about the network interfaces that
the name server might or might not look for. ie The clause doesn't
mean what you'd like it to mean.

    Marten> So for what reason does BIND any periodically interface
    Marten> checks?

Because a host's network interfaces are not necessarily static. They
sometimes get enabled or disabled. Sometimes they go away or get
created: for instance when a dial-up or Wi-Fi link comes on-line.
Mobile hosts often change their IP address. A name server usually
needs to know about these kinds of network events. What if you tell
the name server to only listen on a network interface that doesn't
come alive until after the name server has started? Should the name
server give up because the interface isn't available or keep running
and occasionally check for the interface to come on-line?

Bear in mind too that the name server might want/need to use all of
the host's interfaces for *making* DNS queries, not just the one(s)
that it's accepting queries from. What if a network interface comes up
and the host's default route changes? Shouldn't the server use the new
interface and route for resolving queries?

    Marten> That's a bug in my eyes.

So file a bug report.


More information about the bind-users mailing list