Automatic tunnels

["Christian Huitema" <huitema@windows.microsoft.com>]

During the Vienna meeting, I sensed that there was a split in the WG
constituency between those who like automatic tunneling techniques such
as 6to4 and Teredo because they enable automatic deployment of IPv6, and
those who have an instinctive dislike for these technologies and would
much prefer controlled mechanisms and the orderly deployment of tunnels.
I would like to understand how we can resolve this tension.

My personal analysis is that configured tunnels have a lot of drawbacks,
unless they are "very short", in practice if they are provided by the
very same ISP that provides IPv4 connectivity. This opinion is based on
our collective experience with long tunnels: they require collaboration
of a remote entity, and thus explicit configuration; they don't follow
the natural Internet topology and thus often result in rather poor
transmission times; and they are costly to provide, as someone has to
pay for the transmission to and from the tunnel endpoint.

Automatic tunnels like 6to4 and Teredo have the advantage of being
potentially "very short": the transmission between two transition hosts
follows the IPv4 topology, and is thus as short as it gets; the
transmission between transition and native includes a dog-leg, but that
leg can be as short as the distance to the nearest relay, i.e. the
nearest dual stack router. The typical issues with automatic tunnels are
the stability of the IPv6 address (as stable as the underlying IPv4),
the provision of reverse mappings in the DNS, and the possibility of
attacks on or through the relays, most of which can probably be
mitigated. Another often quoted issue is that people might just be
satisfied with the transition technology and never go native, but I
don't believe that this creates much of a difference between configured
and automatic tunnels.

An obvious compromise would be to use no tunnel at all if the ISP
provides IPv6 connectivity, configured tunnels when provided by the
local ISP, automatic tunneling otherwise. However, this begs one
question. Why would an ISP invest in providing local tunnels, instead of
providing native connectivity faster? If the ISP wants to facilitate the
transition, why would it not just provide a local 6to4 relay for the
exclusive use of its customers?

-- Christian Huitema