[ppml] Policy Proposal: Definition of known ISP and changes to

[William Herrin]

On 8/31/07, Brian Johnson <bjohnson at drtel.com> wrote:
> Isn't there already a prefix set aside for mapping ipv4 into ipv6? If
> there isn't, wouldn't doing this imply that anyone who has ipv4 space
> can simply map this into ipv6 and away you go?

Brian,

The short answer is: sort of but not exactly.

There are several specifications that map existing IPv4 addresses into
various portions of the IPv6 address space for specific purposes. For
example, there is a mapping at the sockets API level that tells the
IPv6 sockets API to use IPv4 instead. Another example is the 6to4
protocol which sets up dynamic tunnels to an IPv6 /48 for each IPv4
address that acts as a tunnel router.

There is, however, no general-purpose mapping which specifies that a
block of IPv4 addresses maps to a range of IPv6 addresses that are
genericly routeable on the IPv6 Internet. I made a proposal to create
such a mapping a few months back (IPv4 to IPv6 Migration Incentive
Address Space) but got shot down hard. I followed up with a proposal
that tweaked 6to4 so that its address space could be dual-purposed to
the task but got shot down on that as well.

The chief issue revolves around the table size in the default-free
zone (the backbone). Each provider-independent prefix consumes a slot
in the very expensive TCAM memory of every single router in the DFZ.
To give a ballpark estimate, your slot costs something in the
neighborhood of $0.10-$0.20 per router and there are somewhere north
of 200,000 affected routers.

Regards,
Bill Herrin

-- 
William D. Herrin                  herrin at dirtside.com  bill at herrin.us
3005 Crane Dr.                        Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004