Re: [Sidrops] question on draft-ietf-sidrops-aspa-verification-04

[Alexander Azimov <[e-mail]>]

Yanun, sorry for the late response.

Your suggestion to ignore the first ASN in the path if the prefix is
received from IXP with its ASN in the path sounds reasonable for me.
It will improve the quality of ingress filtering by IXP members. I will add
it to the todo list for the next version.

ср, 20 мая 2020 г. в 10:53, Guyunan <[e-mail]>:

> Alex,
>
>
>
> Please see inline.
>
>
>
> *From:* Alexander Azimov [mailto:[e-mail]]
> *Sent:* Friday, May 15, 2020 10:33 PM
> *To:* Guyunan <[e-mail]>
> *Cc:* Jay Borkenhagen <[e-mail]>; SIDR Operations WG <
> [e-mail]>
> *Subject:* Re: [Sidrops] question on
> draft-ietf-sidrops-aspa-verification-04
>
>
>
> Jay, Yunan,
>
>
>
> Thank you for the suggestion. The mutual transit sounds good to me too.
>
>
>
> Yunan: All right, so let’s assume in one case, the IXP does not prepend
> its own ASN, meaning the RS and one of its RS clients receives the same
> AS_Path, right? According to Section 5.1 (see below) and Section 5.2 (see
> below), I’m wondering why the detection rules are different for this RS
> (obeying Section 5.1) and this RS client (obeying Section 5.2) regarding
> the same AS_Path.
>
> I think I got your point. You are asking why can't we process prefixes
> from RS likewise prefixes from peers, right?
>
>
>
> Yunan: Right.
>
>
>
> The reason is that I've tried to describe the policy that will be
> applicable in general, so it covers the worst-case scenario when IX places
> it ASN in the path. It possible to say that we can apply one policy if IX
> is present and another policy if it is not, but this will
> introduce additional complexity with operational dependencies on IX
> configuration.
>
>
>
>
> Yunan: So let me try to clear things up here. Do you mean there are
> possibly two cases: 1. IXP ASN is placed in the path 2. IXP ASN is not
> placed in the path? For case 2, the handling is supposed to be the same as
> processing prefixes from peers, right? For case 1, what is the possible
> handing procedure? Removing the IXP ASN first and then treat like prefixes
> from peers?
>
>
>
>
>
> --
>
> Best regards,
>
> Alexander Azimov
>


-- 
Best regards,
Alexander Azimov