Re: [v6ops] I-D Action: draft-ietf-v6ops-unique-ipv6-prefix-per-host-07.txt

[DaeYoung KIM <[e-mail]>]

Sent from my iPhone

> On 17 Aug 2017, at 14:17, Lorenzo Colitti <[e-mail]> wrote:

> The /64 host in my scenario in fact now behaves as a router and distributes /96 prefixes to internal devices. E2E connections don't terminate at the /64 node but at /96 devices.
> 
> You said that, yes. the host gets a /64. The hosts behind it get a /96 each. Which means that the hosts behind it have worse connectivity than it does, because they can't further extend the network (without NAT).

I would' mind that. I'm not a manager of s whole site (of /48 typically), and that's neither my interest nor purview.

All I'm interested would be whether I could recursively apply the same mechanism of 'Unique-IPv6-Prefix-per-Host' internally to my node (perhaps a sophisticated robot or a small IoT island).

(At this point, I'd like to see the draft read '-per-Node' rather than '-per-Host'. If the node terminates connections, then it's a host whereas it's a router if it doesn't only to relay packets to connection-teminating internal devices.)

So, I'm not interested in network connectivity the size of a whole (typically /48) site up to theoretically as many as 2^80 (= 128 - 48) nodes. The application area I'd be interested with my scenario would be in an internal network of tens or hundreds nodes if not thousands. Within my /64 prefix, I can in fact give birth to theoretically as many as 2^32 internal devices (each of a /96 prefix), effectively the size of the whole IPv4 Internet which would be way way more than I'd need.




> /96 prefixes are handed out by a /64 host, not by a network operator.
> 
> This requires that hosts will accept a /96 if the network gives them one. Thus, a network operator that wants to limit address space can simply assign a /96 to a host. Then the host will not be able to give other hosts behind it any /96 prefixes. So if the host wants to extend the network to other hosts it has to use NAT.

Let's make this clear before going any further: I'm not at all against principally distributing /64 prefixes to hosts up to this draft. That far, fine.

All I'm asking is whether I could recursively apply the same (or equivalent) mechanism to my /64 host to incarnate a child network behind my host. My internal devices don't get prefixes directly from a network operator upstream to me, but exclusively from my /64 host. No worry about malicious upstream network operators.