I would agree, perhaps wording it that the upstream lookup reduction is only realized if there is a single cache shared among entities capable of doing upstream lookups independently of each other.
On Nov 15, 2013, at 9:33 AM, JINMEI Tatuya / 神明達哉 <[email protected]> wrote: > At Thu, 7 Nov 2013 06:53:28 +0100, > Daniel Migault <[email protected]> wrote: > >> Thanks for the complete answer. As you mention, this benefits not only the >> end users but also authoritative and resolving servers (especially with >> DNSSEC). > > It's already noted in this thread, but: I'd note that's a benefit only > if the resolving server doesn't unify the same query from multiple > clients while it resolves it. For those implementations (like ISC > BIND 9) that do this unification, prefetching would only increase > (although maybe marginally) external query/responses and only increase > the authoritative server's load. So, this is not an inherent benefit > of prefetch itself, but one for a particular type of implementation. > > I'm not saying this to insist prefetch is a bad idea, though. > Reducing worst-case latency for end clients, (again, perhaps > marginally) is an inherent benefit of prefetching for all types of > resolver implementations, for example. I just wanted to clarify one > minor detail. > > -- > JINMEI, Tatuya -- Brian Somers [email protected]
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ DNSOP mailing list [email protected] https://www.ietf.org/mailman/listinfo/dnsop
