Re: [DNSOP] Call for Adoption draft-wkumari-dnsop-root-loopback

[Nicholas Weaver <[e-mail]>]

Trying to be polite here, but this seems just silly, and the only thing really should be "Don't Bother".


Root latency frankly speaking does not matter.  Lookups to the root themselves should be rare, and the responses have very long TTLs (48 hours!).  So this is clearly optimizing something that needs no optimization.

Lets say your resolver is behind an insanely high latency link with 10s RTTs to everywhere else on the Internet.

The additional latency for lookups to the root will be in the absolute noise compared with the latency for the lookups to the TLDs, or for the TCP RTT latency on all connection setup, or anything else like that.  So why waste the effort optimizing the one thing which doesn't matter?



A far better approach if you actually want to improve resolver performance on high latency links would be to prefetch on expiration: if an element is fetched out of the cache at least X times, when you expire it from the cache immediately trigger a new fetch for that name.

This would have a much better performance benefit than a root zone mirror.

--
Nicholas Weaver                  it is a tale, told by an idiot,
[e-mail]                full of sound and fury,
510-666-2903                                 .signifying nothing
PGP: http://www1.icsi.berkeley.edu/~nweaver/data/nweaver_pub.asc