Session 1: Monday, 14:30-15:30 ICT (UTC +7), November 16, 2020
Agenda bashing and Chairs’ Slides (10 mins)
No chairs slides today due to 1 hour meeting.
RD based ORF [Wei Wang/Aijun Wang] (10 mins)
[Aijun] Ask for WG adoption via slide
[John Scudder]: I could not hear you, but
it is reasonable to ask for WG adoption at this time.
[adjun]: (unable to hear)
[Acee]: There was objection of using RD for policy.
[Keyur]: I still believe RD based mechanism are not
useful. Maybe we can take this on the mailing list
and get another round of feedback.
[John Scudder]: Thank you. We will take this
up on the list.
BGP SR Policy Extensions to Enable IFIT [Giuseppe Fioccola] (15 mins)
Ketan: Thank you for your responses and update.
I’m not asking for more information to be added to
this draft. All the actual processing is left to
SR module. My suggestion would be to write this in
a spring document, IGP, or PCEP document.
Giuseppe: Maybe we should put this in an existing
Ketan: I would suggest a new document.
Zafar: My question is draft-song-opsawg-ifit- framework. There are certain elements.
Would it not be better to complete the framework?
Giuseppe: There is an framework in opsawg. We
are building 1 block. This document can be adopted
without the framework by other architectures.
Gyan: I had a question. Section 2 refers to the tunnel encaps draft.
Giuseppe: I will update the draft.
Gyan: How it leverage the tunnel encaps may need to be detailed.
BGP Extended Community for Identifying the Target Nodes [Jie Dong] (10 mins)
Jeff Haas: I do find the idea the of extended
community to filter to be
I would agreeable to RT-Constraint.
Jie Dong: We could make the RR filtering optional.
Jeff Haas: If RT-Constrain is ok, then this
was work was agreed to by IDR. Your slide may not be appropriate for this work.
You may not get the restriction you wish.
Jeff Tantsura: +1 in chat window.
Randy Bush: the EBGP identifier is only unique
in a single domain.
Jie Dong: We first considered the intra-AS case,
and when we extended to the E-BGP case then
it must be limited to a AS-Confederation.
Randy: “Can” is “not identical to MUST”
Jie Dong: We will consider the inter-AS further.
Rudiger Volk: Will you slash the inter-AS
case? If you want to continue with inter-AS
case you’d need control of propagation for which usually
no policy primitives are available in bgp implementations.
Jie Dong: I understand the problems with
Rudiger Volk - You could resolve this by using
Large communities instead of extended communities.
Jie Dong: Could this idea be discussed further
on the mail list.
Gyan Mishra: The use case is for BGP flow
Jie Dong: The one use case is flow-specification.
Gyan Mishra: Standard communities or large
communities may work and be less complex.
You may it independent of the VPN overlay.
If you just need to tag you may be able to
use standard communities/large communities.
Jie Dong: I will compare and get back to you later.
Traffic Steering using BGP Flowspec with SRv6 Policy [Yunan Gu] (10 mins)
Yunan: Any question or suggestions
Acee: Is this an informational draft?
Yunan: It could be an informational draft.
It is a usage.
John Scudder: Informational
Yunan: Is a BCP more appropriate than an
Jeff: It does not fit best current practice.
[5 minutes for switching]
Session 2: Friday, 16:00-18:00 ICT (UTC +7), November 20, 2020
Sue: Need reviewers for BGP-LS documents.
Sue: BGP YANG model need to move forward.
Jeff Haas (from jabber): BGP yang module still needs additional thorough review. We’re still finding bugs.
Sue: 5575bis and flowspec v2 close to publication, some nits to solve. Time to start Flowspecv2.
Adrian: talk about BGP-LS registry rule update.
Alvaro: RFC 7370 gives the guidelines. One thing is allow people to request code points without a WG document. For that the rules in 7370 could be modified.
Ketan: +1, I am hoping a “point fix” RFC may be faster IMHO?
Sue: autoconf design team need to restart, suggest to have an interim in December.
Jie: Regarding the change in segement types, there is
a draft on the te-lsp distribution. We should align theses two drafts.
Ketan: You are correct, we will work on that point.
Dhruv (from jabber): Is Deprecated right word of a codepoint that is actually not yet allocated by IANA?
Sue: WG LC gives people chance to provide comments.
Adrian: Will give a review of the whole 7752bis.
BGP Flow Spec for Tunneled Traffic [Donald Eastlake] (12 mins)
Donald: Ready - Ready for WG LC
Jeff: We have gone through a lot of learning.
You are doing this as new AFI/SAFI sharing.
Since you are leveraging existing coding types,
you are vunerable to one of the extension headaches.
Donald: This draft says for matching the headers that
there is a separate registry. These are all TLV formats.
It does not say anything about unknown types.
Jeff: I see I was looking at version 8 instead of 10.
I will have to review my comment. I will re-review based on
the current draft.
Alvaro: You are using encoding of the … for match.
Alvaro: for chairs, when we are doing 5575bis and flowspec v6, some questions raised about flowspecv2. so which way will be chosen?
Sue: Need WG’s opinion on shall we wrap this with flowspecv2. Are operators waiting for this?
Jeff Haas: the interesting question is whether the encoding will be similar to in flowspecv2? May have this first, then flowspecv2, and merge it.
Sue: show hands on “should this go for flowspec v2”? 4 raised hands, 7 not raise hands. (total 61 participants)
BGP Extensions for Routing Policy Distribution [Huaimo Chen] (10 mins)
John: Any comments?
John: Comments on the list.
John: Any Comments?
Ketan: The concept of primary and backup SL within a CP does not exist for SR Policy. Primary and backup CP exists but is subject to preference based selection. I would recommend authors to progress the base work first in Spring WG before bringing up protocol mechanisms for adoption.
John: This was the same comment.
Acee: If use anycast address, how to locate the UE?
Luay Jalil: Could you clarify which part of the network in the figure you are talking about? clarify which are 5G terms and which are IP domain terms.
Boris: Not sure BGP is the right tool for sending the measurement information.
Linda: The information is the average of the past one period of time, it is not very transit. Will give explanation of the work flow.
Luay Jalil: Not sure whether it is relevant to edge compute.
John (as WG member): How dynamic is the measurement information?
Linda: it is average based.
Stephane: Who is taking the decision based on the information? A controller or a router? What is the impact on BGP existing best path decision process? Need to comply to BGP path selection rules.
Linda: It is the ingress router.
Luay Jalil: 5G is just one use case, better to make the mechanism generic.
No comment from the