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

[Erik Kline <[e-mail]>]

On Tue, Dec 2, 2025 at 11:27 PM Ole Trøan <[e-mail]> wrote:

> Brian,
>
> > On 3 Dec 2025, at 07:59, Brian E Carpenter <[e-mail]>
> wrote:
> >
> > I don't think so. I think Bob's argument is just another version of
> "almost any address range would work for this purpose; it doesn't have to
> be special."
>
> Doesn’t that depend on how we define loopback prefix behavior? I could not
> find that specified in either of the drafts.
>
> The current loopback address is automatically plumbed into existing
> implementations.
> - any traffic destined to the loopback address will be sent to the node
> itself (ie up the host stack). Is that the behavior proposed for the
> loopback prefix? Without any address configuration any packet with a DA of
> loopback prefix is sent up the host stack?
> - routers do not forward any packets with a loopback source. This is hard
> coded in forwarding paths.
>
> How do the proponents of this idea intend for this to work? The
> suggestions of using an arbitrary prefix seems to imply people think no
> implementation changes are needed??
>

<no hats>

I too was curious about what specification we have around "loopback
semantics".  draft-smith-v6ops-larger-ipv6-loopback-prefix had a good start
on this, I thought.

As for a prefix, I think we could repeat the same thing we did with SRv6
SID space and take a chunk out of 5f00::/8.  If we think a nod to 127 is in
order, we could allocate 5f7f::/16 for "loopback semantics".

But I think the bulk of any document that makes any prefix recommendation
will want to be focused on what kinds of behavior "loopback" might mean.