Thursday, 9 November 2023, Session III, 15:00 - 16:30 PT

Chairs: Gunter Van De Velde, Ketan Talaulikar
Note Taker: Ketan, Gunter
Slides
Video
Chat-Stream
Meetecho Recording
Introduction
1.1 Administrivia, Agenda Bashing (Chairs, 5 min) [5/90]

1.2 WG Document Status (Chairs, 5 min) [10/90]

WG Documents
2.0 BGP Link-State Shortest Path First (SPF) Routing (Keyur Patel, 10 min) [20/90]
Keyur: Shepherd review was deep and insightful and changes will have some impact on existing implementations

Keyur: Question for chairs. Since we have had changes since the Shepherd review, should we do another WGLC?
Ketan: Yes, since the changes were substantial, we should do another LC.
ketan: hope to hear from all implemtors on impact of propsoed changes
Jie Dong: informed he is working on implementation
Keyur: will be welcoming interop
Acee Lindem: welcomes feedback on the one change that is going to be fairly different about taking out the new subTLV they added, however Ketan sugegsted to keep the links as just topology
Ketan: The point was that there coule be multiple prefixes on a link. The review are observations and comment and it is up to the authors to use the comments or not

3.0 A YANG Model for BGP-LS, BGP-LS-VPN, and BGP-LS-SPF (Mahesh Jethanandani, 20 min) [40/90]
Mahesh: updates involve added support for BGP-LS attributes anmd ability to clear statistics
Mahesh: next steps involve support for Link State NLRIs, add support for BGP-LS, LS-SPF in RIB and adding additional examples

Ketan: Is there a implementation of TLV having Sub-TLVs, for example in OSPF. Question for Acee.
Acee: We did implement TLV/Sub-TLV.
Mahesh: Was this for a variable number of Sub-TLVs?
Acee: Believe we have support. But why do we need to specify the depth of Sub-TLVs?
Ketan: one example is in OSPFv2/3 extended LSA with L2 bundle member with all the attributes underneat
Acee: doubts if that was done in a simple way. mostlikely needs seperate grouping
Ketan: When introducing new TLV which could have new subTLVs can it be supported
Acee: mostlikely we cold, but it must be flattened
Mahesh: wondering about adding a leafref to yourself

Individual Documents
4.0 Usage of BGP-LS-SPF in Multi-segment SD-WAN (Hang Shi, 20 min) [60/90]

Keyur: Interersting work because now in overlay will do graph building. Is assumption that underlay will be IP or are tunnels involved
Shi: currently IP or MPLS
Keyur: is it IPsec?
Shi: direct links can be IP
Ketan: your slide on 4 shows some type of tunneling.
Keyur: tunneling helps to avoid packet forwarding loops
Ketan: The topology shown is BGP peerings and it is over tunnels and ghe use-case requires that a graph be build and the best or most optuimal path selected to achieve a certain thing
Ketan: The use case goes beyond the DC, but it is still link state topology   
being build and BGP.
Ketan: propose to use link affinities instead of Link type. it has certain benefits due to abstract notion and is used in other applications
Jie: may want to use some more complex policies the color may not be sufficient
Keyur: This new proposed link type will be very BGP-SPF specific SAFI and it would be nice to make a note of that
Shi: WG adoption possible?
Ketan: This work needs some discussion on the list First
Gunter: This work is not part of a charter at this moment. worthwhile to look at this when discussing the LSVR charter updates


5.0 Proposed Update to BGP Link-State SPF NLRI Selection Rules (Jie Dong, 10 min) [70/90]

Acee: about slide3 the behaviour should be more to what the draft means then to what ot literally ways. when the link between R2-R3 is disconnected, then they are no longer directly connected and then rule 1 would not apply
Acee: it could be clarified on the draft to solve thsi problem observed
Ketan: if problems like this are found, then please share on the list and discussing
Jie: problem scenario2: redundant advertisement (slide4)
Jie: problem scenario3: non-deterministic selection result (slide5)
Jie: propose updates to NLRI selection Rules
Jie: what to do as next steps? merge with base document or seperate document?
Keyur: two comments: (1) prob best to keep the document seperate and we believe that draft v28 is ready and stable and (2) your rule number 4 the assumption is iBGP, but that must maybe be more explicit
Acee: The first rule is something that should be fixed in teh base draft
Acee: all other suggestions is because we do not want to change the base BGP advertisement
Jie: udating the selection may be additional burden to sender and receiver
Ketan: When update on base draft is out, please review and flag/identify these items
Ketan: if all behavior is local to the router then this can be in seperate draft, if its not local, then its worthwhile to include in the base spec
Keyur: its a good document, but preferto keep this as seperate work


Closing
6.0 Charter Update Discussion (Open Mic, 20 mins)
Gunter: work started long ago and now we are almost ready to complete our first deliverable bgp-SPF
Gunter: How to handle new applications spaces and want to make sure that there ebough energy in the WG to keep working on new and novel deliverables
Keyur: Work is interresting and the WG should be rechartered
Keyur: The work needs to be expedited. it all took too much time until now. push tyeh draft even if it is to experimentl track
Jim (rtg-ad): agree that work needs to be expedited
Jim: the WG can use virtual meetings instead of face2face meetings
Randy: When do we expect to see draft charter?
Ketan: does the WG agree agree that these are the items of interrest to the WG? the topics have to come from the WG
Randy: We have been discussing this for years
Gunter: the situation is that we now have a shepherd review we wanted to see as pre-req for reschartering
Alvaro: some work on the slide mentioned is done, like L3DL. Maybe because we are so close we should go with protocol maintenance and L3DL and no more wait anymore.
Randy: can i ask AI on new draft charter
Ketan: give a few weeks and LSVR chairs will push something
Randy: if you need help, let WG know
Jim (rtg-ad): the change on the charter for at least the L3DL work seems minimal and lets push that out asap and add more in a later phase