<div dir="ltr"><div class="gmail_default" style="font-size:small"><br></div><div class="gmail_extra"><br><div class="gmail_quote">On Mon, Mar 6, 2017 at 1:41 PM, Stephane Bortzmeyer <span dir="ltr"><<a href="mailto:bortzmeyer@nic.fr" target="_blank">bortzmeyer@nic.fr</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On Mon, Mar 06, 2017 at 10:31:10AM -0500,<br>
Phillip Hallam-Baker <<a href="mailto:phill@hallambaker.com">phill@hallambaker.com</a>> wrote<br>
<span class=""> a message of 207 lines which said:<br>
<br>
> a DNS over UDP resolver is entirely stateless<br>
<br>
</span>You cannot be serious here? How do you handle the fact that the<br>
authoritative name servers do not reply instantly?<br>
</blockquote></div><br></div><div class="gmail_extra"><div class="gmail_default" style="font-size:small">The request is one UDP packet. It is received, if it can be answered from cached data, a response is given. Otherwise a request is made to the second tier and that request is on hold until tier2 responds.</div><div class="gmail_default" style="font-size:small"><br></div><div class="gmail_default" style="font-size:small">That is the only point at which state is required in any part of the front tier and it is only maintained for the pendancy of the tier 2 request.</div><div class="gmail_default" style="font-size:small"><br></div><div class="gmail_default" style="font-size:small">If the system is using TCP then the entire tier 1 network architecture needs to guarantee that the host that processes the first packet in a request will process all the remaining packets. That is quite obviously a major amount of state.</div><br></div><div class="gmail_extra"><br></div><div class="gmail_extra"><br></div><div class="gmail_extra"><br></div><div class="gmail_extra"><br></div></div>