Lame delegations are a scourge, like untreated venereal disease; often accompanied by a similar refusal to acknowledge externalities... or that there's any problem at all: works for me! But the DNS community, or bright people trying to do other things with DNS (I'm one of them), often bring it on themselves.
For instance let's just go down the "until they find a server that isn't lame" rabbit hole: let's look at a server which isn't actually authoritative for the domain it serves. Maybe they're all lame, sorry.
It could be a recursive, and it should be obvious that AA isn't present in the response. But in the real world, people with their own root are forced to lie in order to publish data in the public (root) namespace. This is baked into server configs as forwarding zones for those too lazy to recompile BIND to always return AA. Not that anybody checks, the DNS still works, so why bother? But the root (ha ha, pun!) of the problem is the obsession with One True Root which is staunchly defended by ICANN, kind of like the petrodollar.
Or it could be authoritative, but not for the zone which it is properly delegated. Take my ISP, please. They impersonate the people who delegate their /17 in reverse DNS; I assume it's because they're too lazy or incompetent to manage 128 /24s. Fun fact: sometimes recursives subsisting on this diet of excrement report that they ARE authoritative for the delegator's /16! The delegator doesn't care, they block email reports from their own /17 so no worries about anyone cleaning up after GG Allin finishes his show.
But the DNS authoritahs bring it on themselves: "Additionally, the queries to lame delegated nameservers represent additional load on the network, and at the incorrectly delegated nameserver. For example, incorrectly configured domains represent 12% of queries to GoDaddy nameservers: one out of eight queries received by a GoDaddy nameserver is a lame query, a query for which the GoDaddy nameserver is not authoritative." Cry me a river. Qname minimization can effectively double the number of (unprimed) queries required to resolve a name. Coincidentally, qname minimization issues requests which are intended to fail (hoping for SOA) and NS requests which are unnecessary under the archaic recursion algorithm which they vigorously defend by farting in a different direction. Needless to say, qname minimization doesn't work well with lame servers.
Lame delegations are a scourge, like untreated venereal disease; often accompanied by a similar refusal to acknowledge externalities... or that there's any problem at all: works for me! But the DNS community, or bright people trying to do other things with DNS (I'm one of them), often bring it on themselves.
For instance let's just go down the "until they find a server that isn't lame" rabbit hole: let's look at a server which isn't actually authoritative for the domain it serves. Maybe they're all lame, sorry.
It could be a recursive, and it should be obvious that AA isn't present in the response. But in the real world, people with their own root are forced to lie in order to publish data in the public (root) namespace. This is baked into server configs as forwarding zones for those too lazy to recompile BIND to always return AA. Not that anybody checks, the DNS still works, so why bother? But the root (ha ha, pun!) of the problem is the obsession with One True Root which is staunchly defended by ICANN, kind of like the petrodollar.
Or it could be authoritative, but not for the zone which it is properly delegated. Take my ISP, please. They impersonate the people who delegate their /17 in reverse DNS; I assume it's because they're too lazy or incompetent to manage 128 /24s. Fun fact: sometimes recursives subsisting on this diet of excrement report that they ARE authoritative for the delegator's /16! The delegator doesn't care, they block email reports from their own /17 so no worries about anyone cleaning up after GG Allin finishes his show.
But the DNS authoritahs bring it on themselves: "Additionally, the queries to lame delegated nameservers represent additional load on the network, and at the incorrectly delegated nameserver. For example, incorrectly configured domains represent 12% of queries to GoDaddy nameservers: one out of eight queries received by a GoDaddy nameserver is a lame query, a query for which the GoDaddy nameserver is not authoritative." Cry me a river. Qname minimization can effectively double the number of (unprimed) queries required to resolve a name. Coincidentally, qname minimization issues requests which are intended to fail (hoping for SOA) and NS requests which are unnecessary under the archaic recursion algorithm which they vigorously defend by farting in a different direction. Needless to say, qname minimization doesn't work well with lame servers.
I could go on, but I won't.