Re: [dhcwg] Comments on draft-ietf-dhc-addr-notification-00

[Jen Linkova <[e-mail]>]

On Thu, Aug 10, 2023 at 9:35 AM Michael Richardson
<[e-mail]> wrote:
> In the absense of this protocol the network would have no idea if the address
> was still in use, if just using RAs, right?

Practically, if the address is *in use* (== used for bidirectional
communication), then at least one router would have a reachable or
stale NC entry for that address.
If the address is only used for *sending* traffic, the network shall
either have a form of SAVI/address tracking on L2 or have other forms
of monitoring (e.g. netflow...).
If the address is valid but not used for sending/receiving traffic
(e.g. it's deprecated), the network wouldn't know, unless smart L2
switches perform address tracking.
Anyway I believe it might be slightly out of scope of this draft and this WG...

> If we are using this protocol to populate DNS, or just audit logs, because
> temporary addresses, and the like, then maybe that's okay.
>
> It's starting to feel like maybe this would also be keeping NCE entries hot.
> (I don't object to that, and actually it's a good use, but I don't think the
> ID calls this out)
> I think that it probably keeps the NCE entry hot because the first hop router
> sees traffic from that address as long it's configured.

Traffic *from* the address doesn't necessarily create the NC entry
(hence we had to write RFC9131) - if we are talking about on-link /64
and not draft-ietf-v6ops-dhcp-pd-per-device.
However if the DHCP server supports replies, the router in the return
path would have the entry for a while.

TBH I'm not sure it's necessary such a good thing..Addresses might
stay valid for a long time, and normally deprecated addresses do not
consume network resources. Well, might be another  reason to use
draft-ietf-v6ops-dhcp-pd-per-device ;)

-- 
SY, Jen Linkova aka Furry