[dns-operations] co-operative cacheing

John Kristoff jtk at ultradns.net
Wed May 14 14:52:19 UTC 2008


On Thu, 08 May 2008 14:20:13 +0000
Paul Vixie <paul at vix.com> wrote:

> i think a local broadcast isn't practical for most topologies whose recursive
> nameservers are geographically diverse.  and there's no parallel to the web
> proxy data sharing case you mentioned -- ICP (RFC 2186) and HTCP (RFC 2756)
> are directed point to point protocols.

This to me seems like it could be solved with the kind of mechanism
that was used in token ring source route networks. You can have the
resolver send out the equivalent of an all routes explorer packet to
the pool you want to essentially share the cache with.  Each goes and
fetches their answer and the first one to return wins, but they each
cache what they got.  You can still do DNSSEC and have the added
benefit that the closest (latency-wise) answer gets chosen for those
geographical/view-based systems.  Not perfect as you may have some
inconsistency, but that might be OK in this sort of simple system I
suspect.

John



More information about the dns-operations mailing list