Re: Charging for IPs

[Tore Anderson via NANOG <nanog () lists nanog org>]

On 07/09/2026 16:12, Lukas Tribus via NANOG wrote:
> We loose interface information in traceroutes if use loopbacks for
> ICMP errors. I'd immagine people have different opinions about whether
> this is a problem or not.

If you see a traceroute where hop N is the loopback of your 
peer/provider's router and hop N+1 is the loopback of your router, you 
would normally have very little difficulty determining your ingress 
interface - it is obviously going to be the one that connects you to 
your peer/provider.

The exception would be LAG and ECMP bundles, in that case you can only 
deduce that it is one of the N interfaces in the bundle. Then again, 
with LAG bundles you will not be able to identify the ingress interface 
from a traceroute even if you are assigning globally unique IPv4/IPv6 
addresses to all your eBGP link networks - you are still at the «one of 
the N in the bundle» precision there.

The only real information loss is with ECMP bundles - you do indeed lose 
visibility of the exact member interface in the bundle the traceroute 
transited, but there is of course no guarantee that the production 
traffic you are troubleshooting came in through that same member as the 
one shown in the traceroute - so for most practical purposes, you're 
still at the «one of the N in the bundle» precision here also.

Bottom line: the information loss from sourcing traceroutes from a 
loopback IP is marginal and of little operational value. The benefit of 
not having to assign globally unique linknets to all eBGP interfaces is 
far greater. YMMV. In any case, as Saku pointed out, thanks to RFC 5837 
you can actually have your cake and eat it too.

Tore

_______________________________________________
NANOG mailing list 
https://lists.nanog.org/archives/list/nanog () lists nanog org/message/ZT43AVCMFSQZ7TQH74L3YOJNFYZ2ZMBF/