> Stephane Bortzmeyer <mailto:bortzme...@nic.fr> > Saturday, February 28, 2015 9:27 AM > [Reflections of the week-end.] > > ... > > Since we always have to manage broken software and network glitches, > DNS programs will have to handle bad clients and servers, anyway. If > so, what is the point of having a better signalling? Is it worth the > cost (in implementation, and WG time)? Shouldn't we focus only on > 5966bis and drop the two others?
no. please do not change recommended initiator behaviour in ways that violate reasonable assumptions of current responders. if the client wants to keep a connection open longer than it takes to send and receive the queries they have on-hand, they should find out whether the server agrees with this incompatible change to the connection management logic described in RFC 1035 4.2.2. -- Paul Vixie
_______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop