Open britram opened 6 years ago
This was probably intended for the case where a naming server does not have an assertion or a shard/zone in range to state something about this name. This could happen if the naming authority cannot sign all its assertions in time. Which should never happen. Or if the naming authority does not want to proof nonexistence, it could just return no assertion available which goes against the paradigm of a secure and verifiable naming system.
The only benefit I see, is that in the above cases the naming server directly responds which frees resources at the recursive resolver instead of waiting for the timeout. But if the recursive resolver is not able to handle these timeouts, it either has to reduce the timeout duration or add more storage.
We know (at the client) when the query timed out.