Re: 6rd - Rapid Deployment on existing IPv4 infrastructures - a new approach

[Iljitsch van Beijnum <iljitsch@muada.com>]

Hi Rémis,

On 12 feb 2008, at 21:27, Rémi Denis-Courmont wrote:

> Lets clear: /64 subnet at the CPE plain simply SUCKS. Badly.

> Next thing is IPv6 NAT, when people try to "cascade" access points/ 
> routers
> within their home network. Like it or not, CPEs MUST get more than / 
> 64, and
> MUST support downstream prefix delegation.

> Last time I checked, Free was not assigned the whole IPv4 address  
> space. I
> assume they don't want to provide the Rapid Deployment service to  
> customers
> of *other* ISPs. So while they do only have a /32 IPv6 prefix, they  
> don't
> need to burn 32 further bits to keep track of each of the IPv4  
> address of
> each of their customers. Storing the 24 lower order bits should be
> sufficient. And this leaves 64-32-24=8 rouable bits per customer.

No, /24 is still too much (only two or three ISPs have that much IPv4  
space) and too little (fills up an entire IPv6 /32) at the same time.  
What should happen here is that the actual allocation size is used. So  
if an ISP has a /19, they use up /56 - /19 = /37 for the customers in  
that /19.