Re: Modified IPv6 to unmodified IPv4

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

On 19-okt-2007, at 13:32, Nathan Ward wrote:

> It is unclear to me why one would not just deploy traditional IPv4  
> NAPT and regular IPv6 - it works with existing protocols, is  
> significantly more simple, and end user hardware is (we are told)  
> being worked on.

The crucial difference is that in that case, the host must have IPv4  
connectivity. Although building a NATed IPv4 network would at first  
seem easier or not much more difficult than building an IPv6 network  
with some gateways, I think the latter is a better way forward,  
especially as we're getting closer to running out of IPv4 addresses.  
As soon as ISPs start NATing their customers, many people will be  
behind double NAT, which makes it impossible to open up ports to make  
peer-to-peer protocols work through a single layer of NAT.

If the ISP hosts the IPv6-IPv4 gateways, then it's possible to have a  
clean, routed path between the gateway and the end-user, even if a  
site has multiple hosts, which pretty much requires NAT for SOHO  
sites for IPv4 today.

Another advantage is that the gatewayed IPv4 connectivity and real  
IPv6 connectivity go hand-in-hand so users always get both access to  
the legacy IPv4 infrastructure and the new IPv6 infrastructure that  
(hopefully) allows peer-to-peer applications and new protocols that  
would require an ALG with NAT. With regular IPv4 NAT it's still an  
extra hurdle to get IPv6 in there so large numbers of users will have  
to live with reduced transparency for some time.