On Thu, Nov 5, 2015 at 5:04 PM, Paul Vixie <[email protected]> wrote: > On Thursday, November 05, 2015 10:26:22 PM Ray Bellis wrote: > > > > > IMHO, if a clarification is needed, it's that a client that depends on > > the order of the RRsets in an answer MUST NOT do so. > > there's an installed base consisting of over a billion stubs and a few > tens of > millions of recursives who will fail to parse your response if the CNAME > comes > after the thing it points to. > > generally speaking, that has to be documented, and has first-mover > advantage. > > -- > P. Vixie > > > To keep things working, I agree that that servers MUST order the answers.
I thought of suggesting that "clients SHOULD accept answers in any order". But on second thought, this increases code complexity, for no real benefit. And we will probably see more low-compute-power devices doing some basic DNS lookups that need to keep code to a minimum. -- Bob Harold
_______________________________________________ DNSOP mailing list [email protected] https://www.ietf.org/mailman/listinfo/dnsop
