Re: [dnssd] [DNSOP] QDCOUNT > 1 (a modest proposal)

[Ray Bellis <[e-mail]>]

On 12/02/2023 12:30, Paul Vixie wrote:

> i'll clarify my previous remarks. i am not proposing multiple rcodes, 
> but rather, multiple qtypes. this enforces the all-qnames-the-same 
> constraint implicitly -- because there is only one qname. in your 
> qdcount>1 design, this constraint must be enforced by the responder, 
> which further ambiguates the rcode=formerr signal. also, this is not 
> similar to what i proposed in EDNS1, it's a much simpler thing which 
> better leverages EDNS0.
> 
> in the normal-for-now case in which the responder doesn't grok the new 
> EDNS option, that option will be ignored and you'll get back a single 
> answer to the q-tuple in the qdcount=1 question section. so in a single 
> round trip you get one answer plus an indication of whether the new 
> option for multiple queries is or isn't available from that responder. 
> so there is no new failure scenario and no wasted round trip.

Paul, I wrote a draft on this, to which I recall you had some f2f input 
when I discussed it with you at one of the Prague IETFs:

draft-bellis-dnsext-multi-qtypes-04

Ray