The Crypto-bros are currently consuming 15GW.
But don't suggest that this is a bad thing or evidence that they might be fanatics unable to see the flaws in their ideological contraption or they will start coinsplaining you.
That said, one of the claims being made for 'New IP' that the ITU is to be peddling is that it allows for shorter addresses for efficiency.
I remember we hit a very similar problem with XML in the early days when people would measure the number of bytes the encoding took up and use that as a talking point as if it meant anything. So we wrote a compression scheme that nobody ever used knowing nobody would ever use it but that it would end the argument.
IPv4 and IPv6 headers are both larger than absolutely necessary. WiFi and other wireless connections are prone to lost packets which in the TCP/IP world are only remediated end to end.
So a small 'header compression' scheme for wireless links and 'constrained devices' that enables local retry might be the answer to this type of objection. It is at any rate a caltrop we should have ready in the bag in case of need.
If 40 bytes matter then so do 20. Rather than limiting device addresses to 8 bits (I have far more than 256 IoT devices in my house) use proper IPv6 addresses, just only send them over the wire once per session.
The MAC address is sufficient to identify the device. All that is needed is a session identifier and whatever bits are needed for link layer reliability.
On Mon, Nov 22, 2021 at 8:08 PM John C Klensin <john-ietf@xxxxxxx> wrote:
--On Tuesday, November 23, 2021 10:31 +1300 Brian E Carpenter
<brian.e.carpenter@xxxxxxxxx> wrote:
> On 23-Nov-21 08:19, Nick Hilliard wrote:
>> John Levine wrote on 22/11/2021 18:50:
>>> But does anyone have any idea what this "report' might be?
>>
>> draft-petrescu-v6ops-ipv6-power-ipv4?
>>
>> The methodology raises questions about whether the results
>> are reliable enough to be quoted.
>
> Regardless of that, it's far from clear that even if those
> results were valid in 2017 that they are still valid today,
> and how they relate to total power consumption, because for
> all I know base stations use 10% less power for IPv6 and core
> routers use 12.5% less. Warren, shall I write a draft raising
> those hypotheses?
>
> How much power is being wasted at this moment by domestic NATs
> and CGN boxes?
And how much power is being wasted by this discussion? :-(
