It is very clear that the best way to actually maintain anonymity is to
use the "tunnelling" solution, i.e. relay you requests through an
anonymizing proxy. This is true for both IPv6 and IPv4. Indeed, the
catch is that the provided anonymity is only as good as the anonymizer.
The business of running anonymizing services is fraught with perils,
such as suboneas, search warrants, etc.; however, mere marketeers and
stealth data base collectors seldom have the power to issue search
warrants.
Short of actual anonymizers, our best friend so far has been PPP, and
the dynamic allocation of dial-up addresses. We loose this if the
address is assigned statically; we loose it whether the static
assignment is for an IPv4 DSL address or for an IPv6 prefix. From a
privacy point of view, it would be nice to complement IPv6's "privacy
address" system with a dynamic prefix allocation. The good part of that
is, it would also make network configuration much more automatic.
Sure, dynamic prefixes will stil reveal information about the network
topology, and hence location. Fact is, that information is probably easy
to get anyhow -- look at clues in trace-route, time-zone indication,
HTTP proxies at network boundaries, network delays, prefered Akamai
server, etc. But disclosing a broad-brush location is a lot more private
than disclosing one's personal address...
By the way, don't we have a problem with the current plans in 3GPP, to
burn a unique identifier inthe mobile phones and to use it as the lower
64 bits of the IPv6 address?
-- Christian Huitema