Skip to main content

Early Review of draft-ietf-idr-bgp-bfd-strict-mode-18
review-ietf-idr-bgp-bfd-strict-mode-18-bgpdir-early-aerts-2026-08-16-00

Request Review of draft-ietf-idr-bgp-bfd-strict-mode
Requested revision No specific revision (document currently at 19)
Type Early Review
Team BGP Directorate (bgpdir)
Deadline 2026-08-17
Requested 2026-08-03
Requested by Susan Hares
Authors Mercia Zheng , Acee Lindem , Jeff Haas , Albert Fu
I-D last updated 2026-08-26 (Latest revision 2026-08-26)
Completed reviews Bgpdir Early review of -18 by Nick Aerts (diff)
Secdir Early review of -19 by Daniel Migault
Opsdir Early review of -19 by Yingzhen Qu
Comments
BGP Dir Needs a careful state machine review with BGP State machine and BFD state machiine. 

Security needs thoughts about attacks based on state machine. 

OPS Dir needs a review of changes for operational issues - especially combinations of BGP peers with BGP peres, BFD, and graceful restart.
Assignment Reviewer Nick Aerts
State Completed
Request Early review on draft-ietf-idr-bgp-bfd-strict-mode by BGP Directorate Assigned
Posted at https://mailarchive.ietf.org/arch/msg/bgpdir/he0TTUjjlOOWZxRAVE5V7uWVdhk
Reviewed revision 18 (document currently at 19)
Result On the right track
Completed 2026-08-16
review-ietf-idr-bgp-bfd-strict-mode-18-bgpdir-early-aerts-2026-08-16-00
Hello,

Thanks for the document.
I have been selected to do a bgp directorate “early” review of this draft.
https://datatracker.ietf.org/doc/draft-ietf-idr-bgp-bfd-strict-mode/18/

Document: draft-ietf-idr-bgp-bfd-strict-mode-18.txt
Reviewer: Nick Aerts
Review Date: 16 August 2026
Status: On the Right Track

Major issues:

BGP FSM State Changes (Section 8):

    When reading through the FSM changes: ActiveDelayOpenBfdUpPending,
    ConnectDelayOpenBfdUpPending and OpenSentBfdUpPending look very similar. In
    all three, OPEN has been sent and received and speaker is waiting for BfdUp
    to send out the KEEPALIVE. Could these be reduced to a single parking state
    (OpenSentBfdUpPending)?

Event 26(KeepAliveMsg) handling in new states (Section 8.3.5 & 8.4.5):

    In Connect (ConnectDelayOpenBfdUpPending) and Active
    (ActiveDelayOpenBfdUpPending) the remote peer could start sending KEEPALIVE
    in response to BfdUp before the local FSM handles the BfdUp event,
    currently the TCP session will be reset.

Starting and Stopping BFD Sessions Associated BGP BFD Strict-Mode (Section 7):

    The BFD RFCs (RFC 5880 and RFC 5882) do not define the hold-down timer.
    Should it be part of this document with its parameters included in this
    draft for the BGP FSM? If the hold-down mechanism is part of this document,
    should the BFD session be setup during the transition from Idle to Connect
    / Active? Likewise should the BFD session be torn down when the FSM
    transitions to Idle? This would then align with the language in RFC 4271
    about when BGP "initializes all BGP resources" / "releases all BGP
    resources". The document should specify interaction with
    'AcceptConnectionsUnconfiguredPeers'. To configure the BFD session BGP
    needs the remote address, which is not available before the FSM is created
    for unconfigured peers.

Stability Considerations (Section 10)

    Paragraph on BFD hold-down implies that this halts the FSM from
    transitioning from OpenConfirm to Established. In OpenConfirm a KEEPALIVE
    message has been sent to the remote peer, so they could transition to
    Established before the local peer. This could result in UPDATE messages
    being received whilst still in the OpenConfirm state (which would reset the
    session).

Minor issues:

BFD Strict-Mode Capability Negotiation (Section 6)

    Should the BGP speaker include the BFD Strict-Mode Capability in its OPEN
    message when BFD is not enabled for the session?

Section 8.3.5

    ActiveDelayOpenBfdUpPending -> The Connect state should have the
    ConnectDelayOpenBfdUpPending sub-state.

Section 8.4.5

    ConnectDelayOpenBfdUpPending-> The Active state should have the
    ActiveDelayOpenBfdUpPending sub-state.

Section 8.5.2 & Section 8.6.2

    Should ConnectRetryCounter be reset to zero or should it be incremented?
    RFC 4271 only resets the ConnectRetryCounter on ManualStart, AutomaticStart
    or ManualStop.

Nits:

Section 1.

    The last paragraphs implies BFD parameters to be exchanged, yet Section 5
    only mentions the BFD Strict-Mode Capability.

Section 8.1

* "prevented from send a KEEPALIVE" → "sending".
* "In the OpenSent upon receiving an OPEN message" -> "In the OpenSent state".

Section 8.3.1

* "the the local system"

Section 8.5.1

* "the the BfdAdminDown"