Re: there _is_ IPv6 NAT - just look for it

[Brian E Carpenter <[e-mail]>]

On 18/03/2014 06:31, Manfredi, Albert E wrote:
> Residential ISPs who are stingy with their prefix, mobile hotspots, and any of a host of new applications, such as in-vehicle communications and IoT in general. Situations in which it is hard to predict what the subnet architecture will look like this early on. Mandating 64-bit IIDs seems like a sure way to either have to request more prefixes more frequently, or what the heck, using a NAT makes this simpler.
> 
> If every household and every IoT platform (e.g. automobile) were given a /48, 

We have 35 trillion of them in the current allocation to IANA, so I don't
see why there's a problem. In any case a /56 would be adequate.

> it seems unlikely (to me, as of today) that IIDs less than 64 bits would ever be needed. The problem is, it seems even more unlikely that every such platform is going to be given a /48. So if 64-bit IIDs are going to become mandatory, the only option for future growth is going to be the NAT.
> 
> Too much depends on how these IoT systems evolve. 

Indeed it does. Personally I'd like my domestic system to be very solidly
isolated from the Internet, probably by an application layer entity.
That doesn't call for NAT, even though it might call for ULA.

   Brian

> For example, are refrigerators, electrical breaker boxes, and home HVAC systems going to be requesting /56 prefixes? I’d say, each of these would definitely need prefixes </64. So individual households have to be given something like /48 right off the bat, not just /56. Is this likely? If we guess wrong today, with IIDs fixed at 64 bits, the only options are renumbering or NATs.
> 
> Bert
> 
> 
> From: Lorenzo Colitti [mailto:[e-mail]]
> Sent: Sunday, March 16, 2014 10:00 PM
> To: Manfredi, Albert E
> Cc: Brian E Carpenter; [e-mail]
> Subject: Re: there _is_ IPv6 NAT - just look for it
> 
> On Mon, Mar 17, 2014 at 6:52 AM, Manfredi, Albert E <[e-mail]<mailto:[e-mail]>> wrote:
> The IETF needs to deal with reality, though. I see a 64-bit IID mandate, the existence of /64 prefix assignments to customers, and opposition to IPv6 NAT as being internally inconsistent. Something has to give. I'm starting to see that IPv6 NAT will become popular in short order.
> 
> In what scenario do you see this as being necessary?
> 
>   *   For residential ISPs, there's DHCPv6 PD (and all ISPs I'm aware of are handing out larger than /64, either by default, or if the customer CPE advertises in a DHCPv6 PD hint that it would like larger than /64).
>   *   For pre-rel10 3GPP links with a fixed allocation size of /64, there's draft-ietf-v6ops-64share
> Anything else?