Due to lack of a WG list, sent to dnsop and dnsext
I've updated two drafts: https://datatracker.ietf.org/doc/draft-wouters-edns-tcp-keepalive/ By popular request, the TIMEOUT value can now be used by both clients and servers to manage their resources and expectations. Various people requested we should not be too strict on anycast servers, and we no longer require them to only use a TIMEOUT of 0. Addressed comments from Ray, Mark, Tatuya, and others, and findings of me as a result of testing using a patched dig against common DNS servers capability to keep TCP sessions open already. Reference the problem of 5966 recommending closing idle TCP sessions in seconds. https://datatracker.ietf.org/doc/draft-wouters-edns-chain-query/ (formerly draft-wouters-edns-tcp-chain-query) Most importantly, it no longer requires TCP, but will allow any "source ip validated" transport, and references draft-eastlake-cookies for UDP. Addresses Marc's comment regarding DNS requests > 512 bytes Clarifies why we want NS records (in case we switch from forwarder to standalone recursive) Update on justification why this is better than rapid-fire UDP. Paul _______________________________________________ DNSOP mailing list [email protected] https://www.ietf.org/mailman/listinfo/dnsop
