Re: draft-gont-6man-managing-privacy-extensions-00.txt

[Thomas Narten <[e-mail]>]

Mark Smith <[e-mail]> writes:

> I also think there is a fundamentally incorrect assumption is being
> made - that IPv6 addresses and humans are tightly coupled.

Actually, if you look at trends, they are increasingly tightly
coupled.

Internet access by humans is increasingly through single-owner devices
(recently laptops, but increasingly smart phones and the like, which
are typically single-owner). Indeed, I've seen projections that assert
that sometime this year more people will access the Internet though
Mobile Wireless devices than through traditional computers. The
primary method of Internet access is changing...

> An IPv6 address identifies an end-node, and the traffic to or from
> it, but does not always identify the human that caused that traffic
> to occur.

True, but if it correctly identifies the human in 70, 80, or 90% of
the cases, that's likely good enough for "innovation" to move in and
exploit such characteristics.

Up levelling, this draft simply reopens a discussion that was had back
when stateless addrconf was first defined. Today, there is no easy way
for the network adminstrator to "configure" clients. When stateless
addrconf is used, there are pretty much no mechanisms for
communicating network preferences or policy.

Allowing/disallowing temporary addresses is one instance of a broader
problem. Another example of policy that would be useful to commnicate
is address selection preferences. Hence, if we are going to define
something like an "H bit", we should really look at the more general
problem.

And, even if there was a mechanism for the network to communicate its
preferences/policies, we'd still have the problem that end hosts could
simply ignore or override those preferences. 

There is no good/easy/simple answer here folks.

Thomas