In message <[EMAIL PROTECTED]>, Jeremy Jackson writes:
> or how about using an NS that returns ICMP errors instead of NXDOMAIN,
> perhaps using anycast for reducing network load?
ICMP is not particularly useful unless the nameserver uses
connected sockets. Now that randomised ports are used this
well may be true but there are still lots of nameservers that
don't see the ICMP message even it makes it past the firewalls.
> Would that stop the timeout errors? server is still lame, you just know
> faster?
>
> On Mon, 2008-11-17 at 05:15 +0530, Suresh Ramasubramanian wrote:
> > On Mon, Nov 17, 2008 at 4:20 AM, James Hess <[EMAIL PROTECTED]> wrote:
> > > One of the secondary/tertiary recursive resolvers may hand the client
> > > a cached response that had been obtained before the registrar took any
> > > action.
> >
> > Yes, and that'd make a good case for the good old ops practice of
> > dialing down the TTL for a while before any NS change is made.
> >
> > --srs
> >
> --
> Jeremy Jackson
> Coplanar Networks
> (519)489-4903
> http://www.coplanar.net
> [EMAIL PROTECTED]
>
>
--
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742 INTERNET: [EMAIL PROTECTED]
