Skip to main content

Minutes interim-2023-mpls-10: Thu 14:00

Meeting Minutes Multiprotocol Label Switching (mpls) WG
Date and time 2023-08-10 14:00
Title Minutes interim-2023-mpls-10: Thu 14:00
State Active
Other versions markdown
Last updated 2023-08-16


MPLS MNA Open Calls Series (08/10/2023):

  1. Agenda Bashing

  2. Action Items(Tarek)
    The open and closed AIs can be found at:
    Regarding AI 1:
    [Loa] Do not support the idea, will talk to Mathew.
    Regarding AI 2:
    [Joel] Text proposed, need the WG to make a decesion.

  3. MNA from a BIER perspective(Greg)
    [Tony Li] Was there any consideration given to stickiness in the
    ISD header?
    [Greg] There were some idea, but it will impact the bitstring a
    lot. Jeffery has a solution on how to do PSD based bier.
    [Jeffery] ISD MNA is transparent to bier.
    [Greg] The current solution requires that the MPLS label for bier
    must be at the bottom of stack, is this requirement kept as such or
    is it can be eased up?
    [Loa] This should be discussed between the two WGs(mpls and bier).

    [Greg] If there is an agreement that a new evolved architecture of
    bier over MPLS is needed.
    [Loa] More discussions needed.
    [Jeffery] For Bier, the assumption of the bottom of stack will not
    be channged, the real issue is that what happenns if there is a PSD
    block after the BoS.
    [Jie] Current bier was designed based on MPLS without MNA, now
    with MNA, maybe we can reconsider the old limitations. In addition,
    With ISD MNA, when packets arrive a node, the ISD will be popped
    off, the information carried in ISD will not be conveyed to the
    downstream nodes.
    [Tony P] It's mpls processing until you hit the bier label, if the
    underling preserves the ISD, build a new stack and puts bier on top
    and sent it to next node. The only issue is that the PSD after the
    bier, there are a couple of solution proposals.
    [Jeffery] The BoS MPLS label is just a indicator for the existance
    of a follow-by bier header. We don't carry MPLS informaiton from one
    bier router to another bier router. So the ISD/PSD data will
    terminate at each bier router, they will not be propagated futher.
    [Jie] If the BFRs are directly connected, only one MPLS label is
    needed, where to put ISD? no need for ISD?
    [Jeffery] In this case, do not need ISD at all.
    [Greg] We need to separate underlay MPLS network from bier
    [ICE] Jie's point is that if you want to put ISD with bier and
    want carry the ISD across e2e, you need to remember the context and
    do pop and push at each BFR, it's a bit inconveniency.
    [Greg] SFC does the similar thing. If the ingress BFR sets the
    ISD, the ISD will be automatically replicated as each BFR.
    [Jeffery] Think about the Ethernet and IP, you don't want to carry
    an internet related information from one node to another using
    [Ice] I agree, but it would not work if you want to use MNA to
    carry slice ID across.
    [Greg] If we want the slice id to be traversing the whole domain,
    some additional mechansims needed.
    [Ice] If you do it with PSD, it's probaly easier.
    [Jie] That's one difference between ISD and PSD, PSD can be
    preseved even if the MPLS stack is popped.
    [Greg] We need to discuss it further, more considersations needed.

    [Tarek] Can we state that bier is an action on LSRs?
    [Jeffery] BFR is LER instead of LSR, and bier is the payload of
    [Tarek] LER/LSR is not the core of the disusion. If you're doing
    multicast transport, you know you will do replication at multiple
    nodes, and you know you are going over MPLS, I'd like to do this
    actioin at selected nodes. So the question is: can MNA inovke these
    [Jeffery] ?
    Contiue on the discussion about the PS NMA in Bier, three candidate
    proposals at hand for discusison.
    [Jie] Regarding option 1, if we put MNA inside bier header, this
    is specific to bier, it will not be called MPLS.
    [Greg] It's not to put MNA inside bier, it's to put MNA after bier
    [Jie] If so, that will not be a generic MPLS functionality.
    [Greg] Agree, there are some implications and consequences to be
    [Jie] Opitons does not require that the bier mpls label is put at
    the BoS?
    [Greg] Yes, more condiseration and discussion are needed.
    [Jie] Another option would be to put MNA first and then the bier
    header, an indicator in the MNA to indicate the existence of the
    bier header. This can be called bier over MPLS MNA that is different
    from the current bier over MPLS.
    [Jeffery] Yes, we need to discuss further on these solutions. For
    me, option 1 means that bier is no longer the mpls payload, not sure
    about opiton 2. Both options do not handle the legacy bier case,
    both have issues. I have some thoughts on option 3 that had been
    sent to the mailing list for discusion.
    [Loa] Option2 gives up the requirment that bier label must be at
    the BoS, right?
    [Greg] I think so, more considerations and discussions needed.
    [Jag] Is option 2 applicaable to ISD, does it mean that we can put
    ISD anywhere in the stack?
    [Tony P] Option2 will be much harder, it violate the current RFCs.
    The first option is more subtle.
    [Ice] Option2 will be a very costly option.
    [Jeffery] Not sure the use case, this is just a theoretical
    exercise on what if we have a PSD MNA in bier packet, and what do we
    do. More thing needs to be considered, I started a discusion on the
    mailing, welcome more people to join.
    [Greg] Agree, significant work needs to be done if we want a PSD
    based bier.

  4. draft-ietf-mpls-mna-hdr(did not disuss due to time limiation)
    -- chance for authors to ask for clarification
    -- target is to have discussion as needed on 2023-10-17

  5. Future agenda items for the interims
    Chairs / All
    Please tell the chairs what you want to discuss

  6. AOB