IPv6 Routing table will be bloated?

Wed Oct 27 00:11:40 UTC 2010

On Tue, Oct 26, 2010 at 05:48:13PM -0400, Randy Carpenter wrote:
> Someone who Randy didn't attribute wrote:
> > I think APNIC has a policy that defines the minimum IPv6 allocation
> > based on your current IPv4 allocation/usage. This would fix the
> > problem?
> It would be nice as a start, but does not really take into consideration future expansion needs.
> 
> I would think that you could draw some parallels, though.
> 
> Something like:
> 
> v4 /16 ~ v6 /32
> v4 /12 ~ v6 /28
> v4 /8 ~ v6 /24
> 
> I know it we don't want to equate v4 and v6, but it may help as a guideline for the size of the customer base.

I don't think it's a particularly good metric, either, because it doesn't
take into account the "conversion rate" of IPv4 to IPv6 addresses, which is
wildly different in different networks.

Fer instance, $JOB[-1] is a colo/hosting business, with a fair chunk of IPv4
allocated, and the standard IPv6 /32.  I did the initial IPv6 address plan,
and I'm pretty confident in saying that they'll *never* need any more than
that /32 of IPv6, because their business model means that they pack their
/64s relatively (hah!) densely (typically there's at least one /24 of IPv4
per /64 of IPv6).  However, anyone doing network access is likely to be
replacing an IPv4 /32 with an IPv6 /48, which results in a lot more address
space usage.

Direct conversion between IPv4 and IPv6 will either result in many places
being starved of IPv6 (very bad, as the OP of this thread pointed out), or
space will be massively overallocated (also, not real hot).

- Matt