Mark Andrews wrote:
>> At 14:24 20/08/2008, Paul Vixie wrote:
>>>>> i propose that an empty answer be sent
>>>> What can stand as an empty answer? REFUSED?

>> RCODE=REFUSED seems to me to be a cleaner solution and as nameserver
>> are allowed to send that back any time it likes, this is not even a
>> hit of protocol change.
>> Someone can just write a short document
>> "QTYPE=ANY considered harmful, just REFUSE them" :-)
>> to document the practice and point the people at that complain.

> We have load balancers that send REFUSED rather than NOERROR
> for AAAA queries today. These cause operational problems.

It seems to me that we have the classic problem of responses written for
servers based on certain conditions leaving the clients to try and
figure out what that response really meant. What is really needed is a
document that says the response X will happen under the following list
of circumstances, and response Y will happen under these other
circumstances. Then it becomes easy to figure out what is the most
appropriate response based on some new circumstances. Clients need to
understand the meaning of the response within the context of the query.


to unsubscribe send a message to with
the word 'unsubscribe' in a single line as the message text body.