RE: [EXTERNAL] Re: Extending a /64 (ATN/IPS worked example)

["Templin (US), Fred L" <[e-mail]>]

Bert,

> -----Original Message-----
> From: Manfredi (US), Albert E
> Sent: Tuesday, November 17, 2020 3:03 PM
> To: Templin (US), Fred L <[e-mail]>; [e-mail]
> Subject: RE: [EXTERNAL] Re: Extending a /64 (ATN/IPS worked example)
> 
> -----Original Message-----
> From: Templin (US), Fred L
> 
> > Bert, I appreciate the perspectives but we will want to use robust IID generation
> capabilities such as RFC4941(bis) and RFC7217 without having to set the 24-bit ID
> in each host's address.
> 
> This would only be for the front of the airplane, I take it. For passengers, it's not an issue. They can use SLAAC as is. Up front, I'd be
> using something more predictable than SLAAC anyway? Plus, another advantage of this, you would not need to request anything
> special, like a new /10 prefix, dedicated only to airplanes.

Front of the airplane (i.e., safety domain), yes.

> > Plus, the aircraft ID is the moral equivalent of a VIN; it is
> not a MAC address, nor an OUI,
> 
> What is more similar to an auto's VIN than the Organizational Unit Identifier, plus a few more bits, of a MAC address? The MAC
> address identifies the manufacturer and the device itself, at least in principle, as VINs do. For this purpose, you only need to identify
> the aircraft, in a completely predicable way, I take it. The rest of the IID can be less deterministic. For instance, use DHCPv6, with the
> top 24 or 39 bits always the same.
> 
> > The aircraft ID is an identity, pure and simple.
> 
> So is the MAC address.
> 
> > And, if the aircraft can self-generate an MNP knowing only its ID, then it does
> not have to ask the network to generate one for it.
> 
> And now, you are putting the burden of change on well-established global routing schemes, designed to aggregate routes, only for
> this airline example. That's really my point. I'm always in favor of looking for the path of least resistance.

No, no burden on routing protocols - simply an overlay of a few lightweight
BGP routers (could even be VMs in the cloud) doing exactly what BGP is
intended to do and without polluting any routing tables in the underlay.

Fred

> Bert