[Sidrops] Re: ASPA validation of paths received from customers and providers

[Maria Matejka <[e-mail]>]

Randy,

> >> as a node may have multiple local ASs, is what you want really test
> >> at egress?
> > 
> > Theoretically, yes.  Practically, that's relying on the neighbor to do
> > the checks properly, and that's not how BGP is designed to work, imho.
> 
> i am not sure what you're meaning here.  i assume not that i am not
> responsible to ensure that what i send is proper, and solely the
> receiver is responsible for checking it.  i don't think we have
> signaling mechanisms with granularity sufficient to the latter.  and,
> even if we did, dealing with diagnostic signal from receiver to sender
> would seem to be more complex than just being more careful with what we
> send.

The export check on my neighbor's BGP speaker shouldn't be enough for me
to not check the same on import.

> my suggestion is not to have the neighbor report my misconfiguration,
> but for me to detect it before sending to neighbor.  see e.g. RFC 8893.

I want to detect my neighbor's misconfiguration before considering their
route for best route selection (and subsequently failing to export it
further).

Maria

-- 
Maria Matejka (she/her) | BIRD Team Leader | CZ.NIC, z.s.p.o.