On Fri, Sep 30, 2022 at 10:46 PM Brian E Carpenter <brian.e.carpenter@xxxxxxxxx> wrote:
On 01-Oct-22 15:06, Phillip Hallam-Baker wrote:
> The fundamental brokenness in the IPv6 transition strategy is that an IPv4 device cannot reach an IPv6 service.
If you believe that, I suggest starting at RFC 8215 and working backwards. NAT46 has been on the radar for many years.
To make it work, you need to add NAT and a signalling layer which is directly equivalent to the one I sketched out.
In practice, the CDNs can solve this too, as well as the other way round.
It hasn't been much of a real life problem, so far, since most services are IPv4-only or dual stack.
But NAT works fine for that situation as well.
The reason I need extended addressing is so I can run servers from my home network. Which is of course something my ISP doesn't want me to do. So why would they deploy infrastructure to...
Being able to do something without the ISPs cooperation might well be the precondition to getting their cooperation.
Instead of just mapping IPv6 packets to an IPv4 address plus port, maybe mapping them to an IPv4 address, port and prefixed payload would be an interesting NAT 64 strategy. Then the NAT can be stateless.
