'Today' considers data retention and IMP

[Igor Mozolevsky]

2009/1/12 Richard Clayton <richard@highwayman.com>:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> In article <a2b6592c0901120234nb912a8bn72009f8f739ddfa3@mail.gmail.com>,
> Igor Mozolevsky <igor@hybrid-lab.co.uk> writes
>
>>AFAIK, DNS is not context-aware,
>
> It can trivially be made so -- and indeed Akamai's entire business model
> depends on this happening :)   [and doubtless other Content Distribution
> Networks as well, with which I am less familiar]
>
>>so no matter where you type
>>'mail.google.com' you will always get a bunch of addresses from the
>>same address pool.
>
> It's far __less__ likely to be the same bunch of addresses for a service
> such as this, than for some random www.example.com.  Google may well
> wish to load balance on a regional basis -- so that DNS requests from
> Europe get given different sets of IP addresses than those from the USA.
> There's no way of telling short of experimenting.

The reason why I made that assumption is that our DNS server is fed
from the root servers, not ISP's DNS servers, so I don't see how you
can fake a geo-aware response, unless there's someone running a DNS
query interception, or Google's DNS servers reply differently based on
the source address of the packet, which I would imagine would be very
expensive. L4 diversion switches (what CDNs use) are not the same as
geo-aware DNS responses, but that doesn't get us any closer to where
one's data actually lives.

--
Igor :-)