Re: Comments on draft-thain-ipv8-01

We need a policy against AI slop. Well, the whole world needs one, not
just the IETF.

/Carlos

On Thu, Apr 16, 2026 at 11:55 AM Caleb James DeLisle <[email protected]> wrote:
>
> I just started reading it until I got to this:
>
>  > draft-thain-routing-protocols-00 BGP8, IBGP8, OSPF8, IS-IS8, CF
>
> and that was enough.
>
>
> If somebody really wants to use legacy software and their DNS server is
> also their NAT (usually is), they can have it map their IPv6 destination
> to a non-routable IPv4 address. Their software will get an A record with
> a 10.0.0.0/8 or something and when it sends to that address the NAT will
> translate it to the correct IPv6. And this whole arrangement can stay a
> dirty little secret between them and their gateway.
>
>
> Thanks,
>
> Caleb
>
>
> On 16/04/2026 15:41, Andrew wrote:
> > This was sent to me by … maybe a half dozen people.
> >
> > Is this some sort April Fools joke?
> >
> > It seems like, by adopting this whole model, we would gain:
> > - Legacy applications can continue using IPv4 sockets (AF_INET and 32-bit address types) * but only if they use the system DNS resolver? *
> > - We can continue using dot-decimal notation instead of scary hex digits
> >
> > In exchange, we need to force everyone to adopt:
> > - new network stack in every OS
> > - new DNS resolver in the OS which can manage prefixes for legacy applications (which is architecturally incompatible how resolving works on Linux)
> > - significantly higher burdens on embedded devices to speak IP
> > - new incompatible firewalls
> > - new incompatible proxies
> > - new incompatible attributes for any protocol which transfers IP addresses
> > - new routing metric (which has not been used in any protocol before)
> > - new incompatible OSPF
> > - new incompatible IS-IS
> > - new BGP AFI
> > - new DNS record type
> > - new address space for IANA+RIRs to manage
> > - new access control protocol, which is unlike any protocol we have seen before
> > - force the use of DHCP again, even though we finally have almost gotten rid of it
> >
> > Compared to IPv6, even if fully implemented into every OS and client and router, the proposed protocol would still lack sufficient address space -  if each ASN has one 32 bit address space, then an ISP who currently has a  v6 /20 would not have enough space to give customers more than a single address, so we are back to NAT again. It doesn’t look like any sort of prefix delegation is used for customer networks, so I guess it’s back to NAT again anyway?
> >
> > How quickly can we reject this?
> >
> > Andrew
> >
>