Skip to main content

Minutes IETF124: grow
minutes-124-grow-00

Meeting Minutes Global Routing Operations (grow) WG
Date and time 2025-11-06 22:00
Title Minutes IETF124: grow
State Active
Other versions markdown
Last updated 2025-11-23

minutes-124-grow-00

Minutes Grow Thursday

Opening Chair: Job Snijders and Paolo Lucente

Generic Event Notification (GEN) for BGP Monitoring Protocol (BMP), Narasimha Prasad

Presentation of the slides.

Request WG adoption.

Thomas asked how to distinguish a RIB-view unmonitor event from a BMP Peer Down event.

Prasad clarified that Peer Down indicates a peer-specific failure, whereas a GEN unmonitor event is used to report router-wide or view-level changes where the peer is not actually down.

Maxence discussed defining a TLV with peer address for local RIBs, noting IPv6 link-local addresses may require interface identifiers.

Maria suggested replacing STLVs with TLVs, as the peer address alone is insufficient to uniquely identify multiple connections with different configurations.

Common BMP Route-Monitoring Messages for Routes Unchanged by Policy, Narasimha Prasad

Presentation of the slides.

Request WG adoption.

No comments.

Definition for BMP Multiple Peer Header, Narasimha Prasad

Presentation of the slides.

Thomas noted a potential overlap with draft-ietf-grow-bmp-loc-peer, and suggested the working group consider whether to advance both documents or select one.

Maria expressed a preference for fewer, more comprehensive RFCs, making it easier to understand how the documents interconnect, especially for future implementations.

Prasad agreed with Maria. Will continue this discussion and get back to the working group.

Definition for Aggregated BMP Route Monitoring Message, Narasimha Prasad

Presentation of the slides.

Request WG adoption.

Luuk requested that authors and vendors provide example data for both unaggregated and aggregated route monitoring, so implementers of BMP collectors can verify interoperability across different vendor implementations.

Luuk asked why TLVs were not included for BGP updates in the BMP PDU, given the effort to make all message types TLV-capable.

Prasad replied that this could be added and clarified in the draft, showing how it would work with the v3 version.

Paolo agreed with Luuk’s comments and suggested enhancing the presentation with concrete use cases and examples to make the objectives and expected outcomes clearer.

Prasad responded that some of this is already covered in the draft and offered to summarize it for the working group.

Current Options for Securing Global Routing, Tobias Fiebig

Presentation of the slides.

BGP Operations and Security, Tobias Fiebig

Presentation of the slides.

Currently Used Terminology in Global Routing Operations, Tobias Fiebig

Presentation of the slides.

The authors are asking the WG and chairs to start WGLC for draft-ietf-grow-bgpopsecupd.

Daniel expressed support for moving the draft forward to working group last call.

Job (speaking as WG chair): The request for WGLC is noted.

Route Server Next Hop Translation, Daniel Wagner

Presentation of the slides.

Request WG adoption.

Tobias expressed support for adoption of the draft but noted some terminology issues, including the use of “unnumbered” when a IPv6 address is present, and raised concerns about ARP/ADP proxying versus route injection, suggesting these topics be discussed during the draft’s lifetime.

Jeff noted potential issues with using the 240 address range for signing, mentioning that Diana is strongly opposed, though some providers are currently doing so. He suggested exercising caution and also considering the 169.254.x.x DHCP range.

Daniel responded that he had already provided some comments on the topic, noting that the concerns have been raised and considered.

BMP Snapshots, Luuk Hendriks

Presentation of the slides.

Request WG Adoption.

Jeff noted that we want “end of rib” markers for end of snapshot for a given afi/safi. It’s very likely that implementations will be leveraging their route refresh infrastructure already.

Storing BMP messages in MRT Format, Luuk Hendriks

Presentation of the slides.

Jeff emphasized the value of serializing BMP messages for tool use and snapshots. He noted that the snapshot mechanism presented is fine, but cautioned against tightly coupling it with on-the-wire re-sync. He suggested keeping the implementation simple, treating the snapshot like a fresh session, and possibly using a new TLV to indicate snapshot-specific purposes.

Paolo commented that the work is highly relevant and useful, highlighting use cases such as reproducing events for debugging and troubleshooting, sharing traces between operators, and analyzing historical events. He suggested these use cases could be incorporated into the draft text.

Camilo: I would say that I agree that there's an overlap between the files and the snapshots, I mean, we have discussed this briefly because to me, how stuff in directories is like having it's like in a Kafka street and having different topics But it's really easy to have different topics in a file system because it's easy to have different directories So the namespace doesn't match So I will go for that I mean, we can And of course, it's useful where it's in MRT

Job emphasized that BMP data stored in MRT files may be archived for decades, and highlighted the need for robust mechanisms to detect and repair corruption, such as checksums or length fields, citing past issues with large-community data causing archive corruption.

Jeff agreed that archiving snapshots is valuable, but noted that saving “bad packets” in the stream can be risky for route monitoring. He supported using detached checksums to ensure data integrity in the archives.

BMP v4: TLV Support for BGP Monitoring Protocol (BMP) Route Monitoring and Peer Down Messages, Paolo Lucente

Presentation of the slides.

Open Question: Shall we merge draft-younsi-grow-bmp-snts?

Thomas suggested rephrasing the open question, noting that Section 6 of the TLV draft describes how timestamps should be handled, while the current draft adds additional timestamps in the TLV, and recommended choosing one approach.

Paolo agreed, explaining that his initial motivation was minimal (“mandatory timestamps”), but Maxence’s work provides a more complete solution, which aligns with their needs.

Thomas noted that this approach matches what has been implemented in Yang Push.

Paolo added that incorporating this draft into the TLV draft would help align the two approaches.

Luuk Hendriks raised the idea of increasing the size of the flex field for flags instead of a version bump, noting he hasn’t heard further feedback and suggesting the group should decide.

Paolo responded that TLVs allow flexible expansion without doubling current dimensions, but cautioned against overusing flags and unnecessary version bumps.

Luuk noted the high demand for flags and suggested a hybrid solution could be considered.

Paolo agreed that the topic is important and should continue to be discussed.

Using BMP over QUIC connection, Yisong Liu

Presentation of the slides.

Request WG adoption.

No comments.

Peer Capability Update Notification in BGP Monitoring Protocol (BMP), Changwang Lin

Presentation of the slides.

Paolo (speaking as WG chair) noted that the work may overlap with Prasad’s GEN message proposal regarding stopping advertisement within a session. He suggested the authors discuss potential alignment or combination of the two approaches.

Maxence asked whether the capability flag should indicate both send and receive directions, or if separate messages are needed.

Changwang clarified that the BGP dynamic capability draft includes context in the PTO, and the flag here simply indicates send or receive direction.

John cautioned that the BGP dynamic capability draft is not yet stable, so detailed design work might be premature. He suggested any detailed concerns be addressed on the mailing list.

Prasad replied to Paolo’s earlier comment, noting two options: extend the GEN message with a TLV or attach the TLV to the peer itself, and that either approach could be acceptable depending on working group preference.

BMP Sequence Number, Timestamp and Flags TLVs, Maxence Younsi

Presentation of the slides.

Jeff emphasized the importance of clearly defining what each timestamp represents for different event types, noting that operators rely on timestamps for continuity in databases. He suggested surveying router vendors to see if sub-second granularity is commonly used.

Prasad agreed on the need for a generic timestamp TLV and noted that higher granularity increases per-route memory usage, so most implementations likely stick to second-level granularity.

Maxence clarified that the TLV allows exporting any existing timestamp, potentially with multiple TLVs, and does not require supporting all types of timestamps.

Thomas recommended considering RFC 9557 on timestamping to ensure consistency across network telemetry systems.

BMP Loc-RIB: Peer address, Maxence Younsi

Presentation of the slides.

No comments.

EVPN-Specific BMP RIB Statistics Extensions, Dikshit Saumya

Presentation of the slides.

Thomas: I just wanted to comment think this is very useful. like a lot that you're now also going down to the different drought types especially that you're well aligning with uh IETF Young module what's being exported there Thank you very much

Open Mic

Jasper proposed flags for filtered Adj-RIB-In/Out to indicate inclusion/exclusion; Jeff noted complexity of expressing full policy, suggesting policy name; Camilo noted YANG module exposes filters for configuration visibility but cannot fully capture all policy details.

Jeff noted that summarizing complex filtering policies in the protocol is difficult, as customer policies vary widely. The best practical approach may be to provide a policy name and let the receiver interpret it.

Camilo added that the YANG module attempts to expose these filters for configuration visibility, but acknowledged the challenges in fully capturing all policy details.