[sig-policy] Re: [apnic-announce] COMMNETS REQUIRED ON ASSIGNMENT OF IPv
We have been running an experimentaio IPv6 exchange, called NSPIXP6,
in Tokyo since Aug 1999. More than 20 IPv6 ISPs are connected
together. See http://www.wide.ad.jp/nspixp6 for detail.
In IPv4 IXes, the addresses of an IX are used only for BGP peering.
In most cases, the IX addresses are not advertized as NextHop field of
IBGP updates using "next-hop-self". Therefore, provided if no user
tries traceroute, the address of an IX could be one of the private
address space.
In IPv6, we have link-local address. In the specification, it is
possible to have a BGP4+ peering using link-local addresses. If a user
performs traceroute with a global source address, the router at an IX
w/ link-local address would repond with one of its global addresses (of
other interface).
Unfortunately, the number of the implementations which can correctly
handle such configuration is very small at the moment; at least zebra
(see http://www.zebra.org) works well. It is expected that other
implementation will support this soon.
So when most of the commercial routers support above configuration, we
may not need to assign a global address prefix to an IPv6 IX.
Small provider independent IPv6 addresses might be necessary to those
systems which offer global services, in which depending one or a few
specific (s)TLA might not be suitable. Those servers may include DNS
servers serving very high level (root or gtld) zones. But this is not
a scope of this comment solicitation.
-- Akira Kato, WIDE Project
* sig-policy: APNIC SIG on resource management policy *
* To unsubscribe: send "unsubscribe" to sig-policy-request at apnic dot net *