[IPv6]Re: [v6ops] Re: New draft: "The IPv6 Loopback Address Prefix"

[Geoff Huston <[e-mail]>]

[I have restored 6man to this conversation - this is an address architecture topic and has its logical home in 6man, according to the 6man charter.]


> On 26 Nov 2025, at 1:03 pm, Alejandro Acosta <[e-mail]> wrote:
> 
> 
> On 25/11/25 9:23 PM, Maciej Żenczykowski wrote:
> 
> 
> [...]
> 
>> ::/104 (so it matches the 24-bits of 127/8)
> 
> I support this option (::/104), IMHO it is more similar to IPv4 and will ease the -on going- transition.
> 
> Alejandro,
> 
> [...]
> 
> 


If a network operator or host admin wants to make an address plan that uses a subset of 16.7M addresses (or 24 bits) and pull it out of an encompassing 96 bits of loopback address prefix, then absolutely nothing is stopping them. Its not as it there is a pressing use case for those other 8 bits, so I would regard a /104 and a /96 as being equivalently useful prefixes in such a scenario.

So why not go the whole way and designate ::/64 as a loopback address prefix? In my mind the constraint was the prior delegation of 0:0:0:0:0:FFFF/96 as IPv4-mapped addresses So I drafted the proposal using the largest bitmask-aligned prefix that was "below" the IPv4-mapped prefix in the IPv6 address space.

I also observe that, as was the case for the documentation prefixes, if we have got this wrong and its too small as address pool we can always return to this question end designate more address space prefixes as loopback.

regards,


   Geoff