* Randy Bush:
IMHO, fate-sharing as a strategy for increasing availability is somewhat underrated.
from rfc 2182
Randy, I didn't write, "don't keep off-site name servers". I wrote, "keep on-site name servers, even if you pay for off-site name service".
3.3. A Myth Exploded
+ While positive DNS results are usually cached, the lack of a result is not cached. Thus, unnecessary inability to resolve creates an undesirable load on the net.
This has been corrected in some implementations since then.
It is important that there be nameservers able to be queried, available always, for all forward zones.
Not answering crap queries (such as queries to addresses for which the resolver has a good reason to believe that they are still unreachable) tends to increase network load, but in some cases, it's the only way to make people notice the problem (like flooding servers with identical queries at an 1/RTT rate). It pushes some of the hurt to a place where it can be addressed. But looking back at incidents such as the Zonelabs/Abovenet issue, your advice is correct for the network we have today. However, we're really covering up a resolver implementation issue, nothing more.