CentOS - Mar 2013 - dhcpd options

A few weeks ago, suddenly, reading news at lunch, I could not get to
nytimes.com. I could ping it, and nslookup it, and if I put the IP address
in place of the name, it was fine.

After *much* back and forth over a ticket I put in, over the last week or
so, our group figured it out: It *seemed* to be related to IPv6, and
there's only *some* few sites, such as the Times, and Orbits, and one or
two others I found, while, say, the Washington Post, or Huffpo, etc, were
fine; and I could make it work by putting IPV6INIT="no" in ifcfg-eth0
and
restarting the network, but another admin got it nailed, we *think*:
apparently the M$-based DNS resolver's sending back extended DNS packets,
and we gag. tcpdump saw us asking for an A record, then an AAAA record,
then using search....

But putting the *very* counterintuitive option edns0 in /etc/resolv.conf,
it works instantly, no caching, no nuthin'. Taking that out breaks it
again.

My question: it *appears* that we could add
option edns0
to dhcpd.conf on the server, and it would fix it for everyone on our subnet.

Have I misunderstood what I'm reading in the manpages?

        mark

m.roth at 5-cent.us wrote:
> A few weeks ago, suddenly, reading news at lunch, I could not get to
> nytimes.com. I could ping it, and nslookup it, and if I put the IP address
> in place of the name, it was fine.
>
> After *much* back and forth over a ticket I put in, over the last week or
> so, our group figured it out: It *seemed* to be related to IPv6, and
> there's only *some* few sites, such as the Times, and Orbits, and one
or
> two others I found, while, say, the Washington Post, or Huffpo, etc, were
> fine; and I could make it work by putting IPV6INIT="no" in
ifcfg-eth0 and
> restarting the network, but another admin got it nailed, we *think*:
> apparently the M$-based DNS resolver's sending back extended DNS
packets,
> and we gag. tcpdump saw us asking for an A record, then an AAAA record,
> then using search....
>
> But putting the *very* counterintuitive option edns0 in /etc/resolv.conf,
> it works instantly, no caching, no nuthin'. Taking that out breaks it
> again.
>
> My question: it *appears* that we could add
> option edns0
> to dhcpd.conf on the server, and it would fix it for everyone on our
> subnet.
>
> Have I misunderstood what I'm reading in the manpages?
>
Sorry, missed some info: CentOS 6.4