On 2010-02-26 09:40, Joel M. Halpern wrote:

>>    Most of this document does not distinguish between IPv4 and
>>    IPv6. The overall routing archtecture for the two protocols is the
>>    same. Consequently, most of the issues in this document apply
>>    equally to IPv4 and IPv6. Any behavior that places pressure on IPv4
>>    routing is likely to also exert the same pressure on IPv6.
>>    Deployment of IPv6 will not lessen these pressures in most cases.
> What you propose to add is good.  But, the text in 4.6 as written claims
> "It is possible to extrapolate what the size of the IPv6 routing table
> would be if wide spread adoption of IPv6 occurred..."  Unfortunately,
> the extrapolation that then takes place assumes that the same factors
> that currently constrain IPv4 sizes would constrain IPvb6 sizes, and
> that seems extremely unlikely.  Hence, this extrapolation is very
> optimistic, and misleading to the reader.

Joel, as you know, it isn't the length of the announced prefixes
that really matters, but the number of non-aggregated prefixes.
So the question is, what constrains the growth in that number?

That is of course very debatable, and there is a lot of research
literature about it. If current trends persist, for example,
you can conclude that a network of 10 billion hosts would
generate just over one million BGP4 routes.
(That is making a scientifically unjustified extrapolation
of the graphs in my CCR paper). But if the rate of multihoming
increases by a factor ten, and if multihoming continues to
require de-aggregation, we'd get ~10 million routes in BGP.
Numbers beyond that imply a much bigger deviation from current
practice than seems likely to me, since it's clear that the
majority of mobility needs will be met using captive-customer
addresses that are intrinsically PA.

My conclusion is that we can hope for 1 million routes but
should have a solution for 10 million. I don't think there is
anything in the data or technology that tells us we will
need a solution for 100 million or more.

All IMHO of course.

Thomas just wrote:

> NEW:
>    It is possible to extrapolate what the size of the IPv6 Internet
>    routing table might look like today, from the current IPv4 Internet
>    routing table, if widespread IPv6 adoption were to occur
>    "instantaneously",
> Then, at the end of the paragraph add something like:
>    Of course, this estimate is based on a current snapshot of IPv4
>    routing activity. Unless the pressures described elsewhere in this
>    document are reduced, the actual table size would be larger.

I don't think that quite covers the very reasonable speculation
that the Internet will grow to around ten billion publicly-
addressable hosts under IPv6, and that this is part of the
problem. So I think we need to be clear that significantly
larger numbers are in our future.

rrg mailing list

Reply via email to