I expect that DNS over TLS and DNS over HTTP/2 are both going to get to much the same place because the technology is driving to the same place: get more of the query into the initial incoming packet so that the first response has useful payload. Do you think the differences are down to more than implementation variations? If there was an innate difference in speed, thats materially interesting.
I would expect that any kept-alive transport has a cost for re-key, which amortizes out over the specific query. If you had enough queries in the pipe, it might not be much overhead against other costs. (I'm not a cryptographer and I sometimes think I'm not much of a network specialist either) -G On Thu, Mar 10, 2016 at 12:21 PM, Tony Finch <d...@dotat.at> wrote: > Davey Song <songlinj...@gmail.com> wrote: >> >> 1) Keep-alive does reduce latency in long time queries. It is a little >> surprising to see that with keep-alive, DNS over HTTP’s latency is almost >> the same as UDP. > > That's not unexpected on a fast link, but it would be worth estimating the > difference in serialization latency on slow links. > > Tony. > -- > f.anthony.n.finch <d...@dotat.at> http://dotat.at/ > Portland, Plymouth: North 4 or 5, becoming variable 3 or less. Slight or > moderate. Mainly fair. Good. > _______________________________________________ > DNSOP mailing list > DNSOP@ietf.org > https://www.ietf.org/mailman/listinfo/dnsop > _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
