Abhayakara
commented
1 year ago It doesn’t seem like we came to a conclusion. I think if we want to allow this, the right approach is to only allow it for queries on the same name, and to make it optional.
The problem is that in order to not cause harm, we need resolvers that don’t support it to respond with just one answer, rather than an error. But we can’t be sure that all resolvers that might exist in this context would do that. So there is the risk that a resolver might respond with some error, forcing the client to send two more messages, or not respond at all, forcing the client to retry multiple times and then probe.
We could argue that using a resolver that doesn’t support whatever behavior we specify is an operational error. If we don’t require that, I think we have to forbid QDCOUNT>1.
Op zo 5 nov 2023 om 18:47 schreef Esko Dijk @.***>
There was a proposal and long discussion on QDCOUNT > 1 in the thread started here: https://mailarchive.ietf.org/arch/msg/dnssd/OpEsujhoSFgC_dXBU6kTd3gtqR0/
Can we conclude on this discussion, if/how we could include it, or not, in the Advertising Proxy's DNS authoritative server which is used to serve query results, based on SRP-registered services, to clients. I.e. could a client use QDCOUNT > 1 (if we put in some specification effort) or not (we abandon the idea)?
— Reply to this email directly, view it on GitHub https://github.com/dnssd-wg/draft-ietf-dnssd-advertising-proxy/issues/2, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA5USAMW6O5OCUMXT44PSO3YC7GL3AVCNFSM6AAAAAA66PQ5B2VHI2DSMVQWIX3LMV43ASLTON2WKOZRHE3TOOBZGEZDKNQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>
EskoDijk
There was a proposal and long discussion on QDCOUNT > 1 in the thread started here: https://mailarchive.ietf.org/arch/msg/dnssd/OpEsujhoSFgC_dXBU6kTd3gtqR0/
Can we conclude on this discussion, if/how we could include it, or not, in the Advertising Proxy's DNS authoritative server which is used to serve query results, based on SRP-registered services, to clients. I.e. could a client use QDCOUNT > 1 (if we put in some specification effort) or not (we abandon the idea)?