[AusNOG] Google's solution to GeoIP for their Public DNS servers

[Martin Barry]

$quoted_author = "Scott Howard" ;
> 
> Ignoring the possible privacy implications people may claim exist
> here, it's an interesting solution.  The biggest problem I see is that
> it's relying on the caching nameserver to do a fair amount of the work
> as cached responses now rely on a "netmask" returned by the upstream
> server, and future queries that might have otherwise have hit the
> cache will require a new lookup if the netmask doesn't match.

What has become clear in the thread on namedroppers.ops.ietf.org is that:

- This only needs to be implemented by recursive resolvers who believe that
  it will provide "better" results. If the recursor is near the edge it is
  serving it does not need to.

- This only needs to be implemented by authorative servers which will
  actually alter the response based on the source IP. Anyone not doing
  "dns-tricks" does not need to.

Paul Vixie's post summarises these two points:
http://www.ops.ietf.org/lists/namedroppers/namedroppers.2010/msg00110.html

cheers
Marty