[6bone] Roadmap to IPv6 multihoming: no PI addresses.

[Michel Py]

Jørgen,

> Jørgen Hovland wrote:
> I dont really think there is a problem. If everybody used ipv6 instead
> of ipv4 right now, we would probably have less prefixes than today.
> There are around 109327 ipv4 prefixes today, but only around 65000 as
> numbers.

About ~12k ASNs are allocated today.

> Im not sure really how big/the usage of a ipv6 /35 is if you compare
> it to a ipv4 /16.  There are no lir's today with more than 1 real ipv6
> prefix anyway? That would result in something like: number of asn's
> == number of prefixes wouldnt it?

We have made the same analysis, and it's close enough although a little optimistic (some will have both a 2001 prefix and a 3ffe); the issue here is that 64k ASNs are not going to last forever, see:
http://search.ietf.org/internet-drafts/draft-ietf-idr-as4bytes-05.txt


> Since becoming a LIR is not free (at least not in europe/ripe), it would
> restrict itself. Small company's probably wont spend that ammount of
> money just to get a LIR membership. Its not worth it.

I have a problem with a system that favors the rich and the powerful. It is a legitimate demand for any size site to be multihomed, especially the small that uses cheesy technology. A large routing table favors big operators, because they are the only ones that can afford to buy the GSR or M160 that gets the job done. We can do better than this.


> Right now, I recieved an email from ripe. They are implementing a new
> global ipv6 policy. Im not familiar with it, but Im looking forward to
> read it. Maybe somebody knows whats it about?

==> - IPv6 End User Site Assignment Request Form in the RIPE
==>    NCC Service Region (for prefixes shorter than a /48)

I have not seen this very form yet, regardless of its name this is the beginning of PI allocation. (Note that I support the new policy by lack of a better one).

Michel.