RFC 9262




Internet Engineering Task Force (IETF)                    T. Eckert, Ed.
Request for Comments: 9262                                     Futurewei
Category: Standards Track                                       M. Menth
ISSN: 2070-1721                                  University of Tuebingen
                                                              G. Cauchie
                                                                  KOEVOO
                                                            October 2022


     Tree Engineering for Bit Index Explicit Replication (BIER-TE)

Abstract



   This memo describes per-packet stateless strict and loose path
   steered replication and forwarding for "Bit Index Explicit
   Replication" (BIER) packets (RFC 8279); it is called "Tree
   Engineering for Bit Index Explicit Replication" (BIER-TE) and is
   intended to be used as the path steering mechanism for Traffic
   Engineering with BIER.

   BIER-TE introduces a new semantic for "bit positions" (BPs).  These
   BPs indicate adjacencies of the network topology, as opposed to (non-
   TE) BIER in which BPs indicate "Bit-Forwarding Egress Routers"
   (BFERs).  A BIER-TE "packets BitString" therefore indicates the edges
   of the (loop-free) tree across which the packets are forwarded by
   BIER-TE.  BIER-TE can leverage BIER forwarding engines with little
   changes.  Co-existence of BIER and BIER-TE forwarding in the same
   domain is possible -- for example, by using separate BIER
   "subdomains" (SDs).  Except for the optional routed adjacencies,
   BIER-TE does not require a BIER routing underlay and can therefore
   operate without depending on a routing protocol such as the "Interior
   Gateway Protocol" (IGP).

Status of This Memo



   This is an Internet Standards Track document.

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Further information on
   Internet Standards is available in Section 2 of RFC 7841.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   https://www.rfc-editor.org/info/rfc9262.

Copyright Notice



   Copyright (c) 2022 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (https://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.  Code Components extracted from this document must
   include Revised BSD License text as described in Section 4.e of the
   Trust Legal Provisions and are provided without warranty as described
   in the Revised BSD License.

Table of Contents



   1.  Overview
   2.  Introduction
     2.1.  Requirements Language
     2.2.  Basic Examples
     2.3.  BIER-TE Topology and Adjacencies
     2.4.  Relationship to BIER
     2.5.  Accelerated Hardware Forwarding Comparison
   3.  Components
     3.1.  The Multicast Flow Overlay
     3.2.  The BIER-TE Control Plane
       3.2.1.  The BIER-TE Controller
         3.2.1.1.  BIER-TE Topology Discovery and Creation
         3.2.1.2.  Engineered Trees via BitStrings
         3.2.1.3.  Changes in the Network Topology
         3.2.1.4.  Link/Node Failures and Recovery
     3.3.  The BIER-TE Forwarding Plane
     3.4.  The Routing Underlay
     3.5.  Traffic Engineering Considerations
   4.  BIER-TE Forwarding
     4.1.  The BIER-TE Bit Index Forwarding Table (BIFT)
     4.2.  Adjacency Types
       4.2.1.  Forward Connected
       4.2.2.  Forward Routed
       4.2.3.  ECMP
       4.2.4.  Local Decap(sulation)
     4.3.  Encapsulation / Co-existence with BIER
     4.4.  BIER-TE Forwarding Pseudocode
     4.5.  BFR Requirements for BIER-TE Forwarding
   5.  BIER-TE Controller Operational Considerations
     5.1.  Bit Position Assignments
       5.1.1.  P2P Links
       5.1.2.  BFERs
       5.1.3.  Leaf BFERs
       5.1.4.  LANs
       5.1.5.  Hub and Spoke
       5.1.6.  Rings
       5.1.7.  Equal-Cost Multipath (ECMP)
       5.1.8.  Forward Routed Adjacencies
         5.1.8.1.  Reducing Bit Positions
         5.1.8.2.  Supporting Nodes without BIER-TE
       5.1.9.  Reuse of Bit Positions (without DNC)
       5.1.10. Summary of BP Optimizations
     5.2.  Avoiding Duplicates and Loops
       5.2.1.  Loops
       5.2.2.  Duplicates
     5.3.  Managing SIs, Subdomains, and BFR-ids
       5.3.1.  Why SIs and Subdomains?
       5.3.2.  Assigning Bits for the BIER-TE Topology
       5.3.3.  Assigning BFR-ids with BIER-TE
       5.3.4.  Mapping from BFRs to BitStrings with BIER-TE
       5.3.5.  Assigning BFR-ids for BIER-TE
       5.3.6.  Example Bit Allocations
         5.3.6.1.  With BIER
         5.3.6.2.  With BIER-TE
       5.3.7.  Summary
   6.  Security Considerations
   7.  IANA Considerations
   8.  References
     8.1.  Normative References
     8.2.  Informative References
   Appendix A.  BIER-TE and Segment Routing (SR)
   Acknowledgements

   Authors' Addresses



1.  Overview



   "Tree Engineering for Bit Index Explicit Replication" (BIER-TE) is
   based on the (non-TE) BIER architecture, terminology, and packet
   formats as described in [RFC8279] and [RFC8296].  This document
   describes BIER-TE, with the expectation that the reader is familiar
   with these two documents.

   BIER-TE introduces a new semantic for "bit positions" (BPs).  These
   BPs indicate adjacencies of the network topology, as opposed to (non-
   TE) BIER in which BPs indicate "Bit-Forwarding Egress Routers"
   (BFERs).  A BIER-TE "packets BitString" therefore indicates the edges
   of the (loop-free) tree across which the packets are forwarded by
   BIER-TE.  With BIER-TE, the "Bit Index Forwarding Table" (BIFT) of
   each "Bit-Forwarding Router" (BFR) is only populated with BPs that
   are adjacent to the BFR in the BIER-TE topology.  Other BPs are empty
   in the BIFT.  The BFR replicates and forwards BIER packets to
   adjacent BPs that are set in the packets.  BPs are normally also
   cleared upon forwarding to avoid duplicates and loops.

   BIER-TE can leverage BIER forwarding engines with little or no
   changes.  It can also co-exist with BIER forwarding in the same
   domain -- for example, by using separate BIER subdomains.  Except for
   the optional routed adjacencies, BIER-TE does not require a BIER
   routing underlay and can therefore operate without depending on a
   routing protocol such as the "Interior Gateway Protocol" (IGP).

   This document is structured as follows:

   *  Section 2 introduces BIER-TE with two forwarding examples,
      followed by an introduction to the new concepts of the BIER-TE
      (overlay) topology, and finally a summary of the relationship
      between BIER and BIER-TE and a discussion of accelerated hardware
      forwarding.

   *  Section 3 describes the components of the BIER-TE architecture:
      the multicast flow overlay, the BIER-TE layer with the BIER-TE
      control plane (including the BIER-TE controller), the BIER-TE
      forwarding plane, and the routing underlay.

   *  Section 4 specifies the behavior of the BIER-TE forwarding plane
      with the different types of adjacencies and possible variations of
      BIER-TE forwarding pseudocode, and finally the mandatory and
      optional requirements.

   *  Section 5 describes operational considerations for the BIER-TE
      controller, primarily how the BIER-TE controller can optimize the
      use of BPs by using specific types of BIER-TE adjacencies for
      different types of topological situations.  It also describes how
      to assign bits to avoid loops and duplicates (which, in BIER-TE,
      does not come "for free").  Finally, it discusses how "Set
      Identifiers" (SIs), "subdomains" (SDs), and BFR-ids can be managed
      by a BIER-TE controller; examples and a summary are provided.

   *  Appendix A concludes this document; details regarding the
      relationship between BIER-TE and "Segment Routing" (SR) are
      discussed.

   Note that related work [CONSTRAINED-CAST] uses Bloom filters
   [Bloom70] to represent leaves or edges of the intended delivery tree.
   Bloom filters in general can support larger trees/topologies with
   fewer addressing bits than explicit BitStrings, but they introduce
   the heuristic risk of false positives and cannot clear bits in the
   BitStrings during forwarding to avoid loops.  For these reasons,
   BIER-TE, like BIER, uses explicit BitStrings.  Explicit BitStrings as
   used by BIER-TE can also be seen as a special type of Bloom filter,
   and this is how other related work [ICC] describes it.

2.  Introduction



2.1.  Requirements Language



   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
   "OPTIONAL" in this document are to be interpreted as described in
   BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all
   capitals, as shown here.

2.2.  Basic Examples



   BIER-TE forwarding is best introduced with simple examples.  These
   examples use formal terms defined later in this document (Figure 4 in
   Section 4.1), including forward_connected(), forward_routed(), and
   local_decap().

   Consider the simple network in the BIER-TE overview example shown in
   Figure 1, with six BFRs.  p1...p15 are the bit positions used.  All
   BFRs can act as a "Bit-Forwarding Ingress Router" (BFIR); BFR1, BFR3,
   BFR4, and BFR6 can also be BFERs.  "Forward_connected()" is the name
   used for adjacencies that represent subnet adjacencies of the
   network.  "Local_decap()" is the name used for the adjacency that
   decapsulates BIER-TE packets and passes their payload to higher-layer
   processing.

   BIER-TE Topology:

      Diagram:

                       p5    p6
                     --- BFR3 ---
                  p3/    p13     \p7          p15
      BFR1 ---- BFR2              BFR5 ----- BFR6
         p1   p2  p4\    p14     /p10 p11   p12
                     --- BFR4 ---
                       p8    p9

      (simplified) BIER-TE Bit Index Forwarding Tables (BIFTs):

      BFR1:   p1  -> local_decap()
              p2  -> forward_connected() to BFR2

      BFR2:   p1  -> forward_connected() to BFR1
              p5  -> forward_connected() to BFR3
              p8  -> forward_connected() to BFR4

      BFR3:   p3  -> forward_connected() to BFR2
              p7  -> forward_connected() to BFR5
              p13 -> local_decap()

      BFR4:   p4  -> forward_connected() to BFR2
              p10 -> forward_connected() to BFR5
              p14 -> local_decap()

      BFR5:   p6  -> forward_connected() to BFR3
              p9  -> forward_connected() to BFR4
              p12 -> forward_connected() to BFR6

      BFR6:   p11 -> forward_connected() to BFR5
              p15 -> local_decap()

                      Figure 1: BIER-TE Basic Example

   Assume that a packet from BFR1 should be sent via BFR4 to BFR6.  This
   requires a BitString (p2,p8,p10,p12,p15).  When this packet is
   examined by BIER-TE on BFR1, the only bit position from the BitString
   that is also set in the BIFT is p2.  This will cause BFR1 to send the
   only copy of the packet to BFR2.  Similarly, BFR2 will forward to
   BFR4 because of p8, BFR4 to BFR5 because of p10, and BFR5 to BFR6
   because of p12.  p15 finally makes BFR6 receive and decapsulate the
   packet.

   To send a copy to BFR6 via BFR4 and also a copy to BFR3, the
   BitString needs to be (p2,p5,p8,p10,p12,p13,p15).  When this packet
   is examined by BFR2, p5 causes one copy to be sent to BFR3 and p8 one
   copy to BFR4.  When BFR3 receives the packet, p13 will cause it to
   receive and decapsulate the packet.

   If instead the BitString was (p2,p6,p8,p10,p12,p13,p15), the packet
   would be copied by BFR5 towards BFR3 because of p6 instead of being
   copied by BFR2 to BFR3 because of p5 in the prior case.  This
   demonstrates the ability of the BIER-TE topology, as shown in
   Figure 1, to make the traffic pass across any possible path and be
   replicated where desired.

   BIER-TE has various options for minimizing BP assignments, many of
   which are based on out-of-band knowledge about the required multicast
   traffic paths and bandwidth consumption in the network, e.g., from
   predeployment planning.

   Figure 2 shows a modified example, in which Rtr2 and Rtr5 are assumed
   not to support BIER-TE, so traffic has to be unicast encapsulated
   across them.  To explicitly distinguish routed/tunneled forwarding of
   BIER-TE packets from Layer 2 forwarding (forward_connected()), these
   adjacencies are called "forward_routed()" adjacencies.  Otherwise,
   there is no difference in their processing over the aforementioned
   forward_connected() adjacencies.

   In addition, bits are saved in the following example by assuming that
   BFR1 only needs to be a BFIR -- not a BFER or a transit BFR.

   BIER-TE Topology:

      Diagram:

                      p1  p3  p7
                   ....> BFR3 <....       p5
           ........                ........>
      BFR1       (Rtr2)          (Rtr5)      BFR6
           ........                ........> p9
                   ....> BFR4 <....       p6
                      p2  p4  p8

      (simplified) BIER-TE Bit Index Forwarding Tables (BIFTs):

      BFR1:   p1  -> forward_routed() to BFR3
              p2  -> forward_routed() to BFR4

      BFR3:   p3  -> local_decap()
              p5  -> forward_routed() to BFR6

      BFR4:   p4  -> local_decap()
              p6  -> forward_routed() to BFR6

      BFR6:   p7  -> forward_routed() to BFR3
              p8  -> forward_routed() to BFR4
              p9  -> local_decap()

                  Figure 2: BIER-TE Basic Overlay Example

   To send a BIER-TE packet from BFR1 via BFR3 to be received by BFR6,
   the BitString is (p1,p5,p9).  A packet from BFR1 via BFR4 to be
   received by BFR6 uses the BitString (p2,p6,p9).  A packet from BFR1
   to be received by BFR3,BFR4 and from BFR3 to be received by BFR6 uses
   (p1,p2,p3,p4,p5,p9).  A packet from BFR1 to be received by BFR3,BFR4
   and from BFR4 to be received by BFR6 uses (p1,p2,p3,p4,p6,p9).  A
   packet from BFR1 to be received by BFR4, then from BFR4 to be
   received by BFR6, and finally from BFR6 to be received by BFR3, uses
   (p2,p3,p4,p6,p7,p9).  A packet from BFR1 to be received by BFR3, then
   from BFR3 to be received by BFR6, and finally from BFR6 to be
   received by BFR4, uses (p1,p3,p4,p5,p8,p9).

2.3.  BIER-TE Topology and Adjacencies



   The key new component in BIER-TE compared to (non-TE) BIER is the
   BIER-TE topology as introduced through the two examples in
   Section 2.2.  It is used to control where replication can or should
   happen and how to minimize the required number of BPs for
   adjacencies.

   The BIER-TE topology consists of the BIFTs of all the BFRs and can
   also be expressed as a directed graph where the edges are the
   adjacencies between the BFRs labeled with the BP used for the
   adjacency.  Adjacencies are naturally unidirectional.  A BP can be
   reused across multiple adjacencies as long as this does not lead to
   undesired duplicates or loops, as explained in Section 5.2.

   If the BIER-TE topology represents (a subset of) the underlying
   (Layer 2) topology of the network as shown in the first example, this
   may be called an "underlay" BIER-TE topology.  A topology consisting
   only of "forward_routed()" adjacencies as shown in the second example
   may be called an "overlay" BIER-TE topology.  A BIER-TE topology with
   both forward_connected() and forward_routed() adjacencies may be
   called a "hybrid" BIER-TE topology.

2.4.  Relationship to BIER



   BIER-TE is designed so that its forwarding plane is a simple
   extension to the (non-TE) BIER forwarding plane, hence allowing it to
   be added to BIER deployments where it can be beneficial.

   BIER-TE is also intended as an option to expand the BIER architecture
   into deployments where (non-TE) BIER may not be the best fit, such as
   statically provisioned networks that need path steering but do not
   want distributed routing protocols.

   1.  BIER-TE inherits the following aspects from BIER unchanged:

       1.a  The fundamental purpose of per-packet signaled replication
            and delivery via a BitString.

       1.b  The overall architecture, which consists of three layers:
            the flow overlay, the BIER(-TE) layer, and the routing
            underlay.

       1.c  The supported encapsulations [RFC8296].

       1.d  The semantics of all BIER header elements [RFC8296] used by
            the BIER-TE forwarding plane, other than the semantic of the
            BP in the BitString.

       1.e  The BIER forwarding plane, except for how bits have to be
            cleared during replication.

   2.  BIER-TE has the following key changes with respect to BIER:

       2.a  In BIER, bits in the BitString of a BIER packet header
            indicate a BFER, and bits in the BIFT indicate the BIER
            control plane's calculated next hop towards that BFER.  In
            BIER-TE, a bit in the BitString of a BIER packet header
            indicates an adjacency in the BIER-TE topology, and only the
            BFR that is the upstream of that adjacency has its BP
            populated with the adjacency in its BIFT.

       2.b  In BIER, the implied reference options for the core part of
            the BIER layer control plane are the BIER extensions for
            distributed routing protocols.  These include IS-IS and OSPF
            extensions for BIER, as specified in [RFC8401] and
            [RFC8444], respectively.

       2.c  The reference option for the core part of the BIER-TE
            control plane is the BIER-TE controller.  Nevertheless, both
            the BIER and BIER-TE BIFTs' forwarding plane state could
            equally be populated by any mechanism.

       2.d  Assuming the reference options for the control plane, BIER-
            TE replaces in-network autonomous path calculations with
            explicit paths calculated by the BIER-TE controller.

   3.  The following elements/functions described in the BIER
       architecture are not required by the BIER-TE architecture:



       3.a  "Bit Index Routing Tables" (BIRTs) are not required on BFRs
            for BIER-TE when using a BIER-TE controller, because the
            controller can directly populate the BIFTs.  In BIER, BIRTs
            are populated by the distributed routing protocol support
            for BIER, allowing BFRs to populate their BIFTs locally from
            their BIRTs.  Other BIER-TE control plane or management
            plane options may introduce requirements for BIRTs for BIER-
            TE BFRs.

       3.b  The BIER-TE layer forwarding plane does not require BFRs to
            have a unique BP; see Section 5.1.3.  Therefore, BFRs may
            not have a unique BFR-id; see Section 5.3.3.

       3.c  Identification of BFRs by the BIER-TE control plane is
            outside the scope of this specification.  Whereas the BIER
            control plane uses BFR-ids in its BFR-to-BFR signaling, a
            BIER-TE controller may choose any form of identification
            deemed appropriate.

       3.d  BIER-TE forwarding does not require the BFIR-id field of the
            BIER packet header.

   4.  Co-existence of BIER and BIER-TE in the same network requires the
       following:



       4.a  The BIER/BIER-TE packet header needs to allow the addressing
            of both BIER and BIER-TE BIFTs.  Depending on the
            encapsulation option, the same SD may or may not be reusable
            across BIER and BIER-TE.  See Section 4.3.  In either case,
            a packet is always forwarded only end to end via BIER or via
            BIER-TE ("ships in the night" forwarding).

       4.b  BIER-TE deployments will have to assign BFR-ids to BFRs and
            insert them into the BFIR-id field of BIER packet headers,
            as does BIER, whenever the deployment uses (unchanged)
            components developed for BIER that use BFR-ids, such as
            multicast flow overlays or BIER layer control plane
            elements.  See also Section 5.3.3.

2.5.  Accelerated Hardware Forwarding Comparison



   BIER-TE forwarding rules, especially BitString parsing, are designed
   to be as close as possible to those of BIER, with the expectation
   that this eases the programming of BIER-TE forwarding code and/or
   BIER-TE forwarding hardware on platforms supporting BIER.  The
   pseudocode in Section 4.4 shows how existing (non-TE) BIER/BIFT
   forwarding can be modified to support the required BIER-TE forwarding
   functionality (Section 4.5), by using the BIER BIFT's "Forwarding Bit
   Mask" (F-BM): only the clearing of bits to avoid sending duplicate
   packets to a BFR's neighbor is skipped in BIER-TE forwarding, because
   it is not necessary and could not be done when using a BIER F-BM.

   Whether to use BIER or BIER-TE forwarding is simply a choice of the
   mode of the BIFT indicated by the packet (BIER or BIER-TE BIFT).
   This is determined by the BFR configuration for the encapsulation;
   see Section 4.3.

3.  Components

   BIER-TE can be thought of as being composed of the same three layers
   as BIER: the "multicast flow overlay", the "BIER layer", and the
   "routing underlay".  Figure 3 also shows how the BIER layer is
   composed of the "BIER-TE forwarding plane" and the "BIER-TE control
   plane" as represented by the "BIER-TE controller".

                   <------BGP/PIM----->
      |<-IGMP/PIM->  multicast flow   <-PIM/IGMP->|
                        overlay

          BIER-TE  [BIER-TE Controller] <=> [BIER-TE Topology]
          control     ^      ^     ^
          plane      /       |      \   BIER-TE control protocol
                    |        |       |  (e.g., YANG/NETCONF/RESTCONF
                    |        |       |       PCEP/...)
                    v        v       v
    Src -> Rtr1 -> BFIR-----BFR-----BFER -> Rtr2 -> Rcvr

                   |<----------------->|
                 BIER-TE forwarding plane

                   |<- BIER-TE domain->|

                 |<--------------------->|
                     Routing underlay

                       Figure 3: BIER-TE Architecture

3.1.  The Multicast Flow Overlay



   The multicast flow overlay has the same role as that described for
   BIER in [RFC8279], Section 4.3.  See also 3.2.1.2">Section 3.2.1.2.

   When a BIER-TE controller is used, it might also be preferable that
   multicast flow overlay signaling be performed through a central point
   of control.  For BGP-based overlay flow services such as "Multicast
   VPN Using Bit Index Explicit Replication (BIER)" [RFC8556], this can
   be achieved by making the BIER-TE controller operate as a BGP Route
   Reflector [RFC4456] and combining it with signaling through BGP or a
   different protocol for the BIER-TE controller's calculated
   BitStrings.  See Sections 3.2.1.2 and 5.3.4.

3.2.  The BIER-TE Control Plane



   In the (non-TE) BIER architecture [RFC8279], the BIER layer is
   summarized in Section 4.2 of [RFC8279].  This summary includes both
   the functions of the BIER-layer control plane and forwarding plane,
   without using those terms.  Example standardized options for the BIER
   control plane include IS-IS and OSPF extensions for BIER, as
   specified in [RFC8401] and [RFC8444], respectively.

   For BIER-TE, the control plane includes, at a minimum, the following
   functionality.

   1.  BIER-TE topology control: During initial provisioning of the
       network and/or during modifications of its topology and/or
       services, the protocols and/or procedures to establish BIER-TE
       BIFTs:

       1.a  Determine the desired BIER-TE topology for BIER-TE
            subdomains: the adjacencies that are assigned to BPs.
            Topology discovery is discussed in Section 3.2.1.1, and the
            various aspects of the BIER-TE controller's determinations
            regarding the topology are discussed throughout Section 5.

       1.b  Determine the per-BFR BIFT from the BIER-TE topology.  This
            is achieved by simply extracting the adjacencies of the BFR
            from the BIER-TE topology and populating the BFR's BIFT with
            them.

       1.c  Optionally assign BFR-ids to BFIRs for later insertion into
            BIER headers on BFIRs as BFIR-ids.  Alternatively, BFIR-ids
            in BIER packet headers may be managed solely by the flow
            overlay layer and/or be unused.  This is discussed in
            Section 5.3.3.

       1.d  Install/update the BIFTs into the BFRs and, optionally, BFR-
            ids into BFIRs.  This is discussed in Section 3.2.1.1.

   2.  BIER-TE tree control: During network operations, protocols and/or
       procedures to support creation/change/removal of overlay flows on
       BFIRs:

       2.a  Process the BIER-TE requirements for the multicast overlay
            flow: BFIRs and BFERs of the flow as well as policies for
            the path selection of the flow.  This is discussed in
            Section 3.5.

       2.b  Determine the BitStrings and, optionally, entropy.
            BitStrings are discussed in Sections 3.2.1.2, 3.5, and
            5.3.4.  Entropy is discussed in Section 4.2.3.

       2.c  Install state on the BFIR to impose the desired BIER packet
            header(s) for packets of the overlay flow.  Different
            aspects of this point, as well as the next point, are
            discussed throughout Section 3.2.1 and in Section 4.3.  The
            main component responsible for these two points is the
            multicast flow overlay (Section 3.1), which is
            architecturally inherited from BIER.

       2.d  Install the necessary state on the BFERs to decapsulate the
            BIER packet header and properly dispatch its payload.

3.2.1.  The BIER-TE Controller



   This architecture describes the BIER-TE control plane, as shown in
   Figure 3, as consisting of:

   *  A BIER-TE controller.

   *  BFR data models and protocols to communicate between the
      controller and BFRs in support of BIER-TE topology control (see
      the list under "BIER-TE topology control"), such as YANG/NETCONF/
      RESTCONF [RFC7950] [RFC6241] [RFC8040].

   *  BFR data models and protocols to communicate between the
      controller and BFIRs in support of BIER-TE tree control (see
      Section 3.2, point 2.), such as BIER-TE extensions for [RFC5440].

   The single, centralized BIER-TE controller is used in this document
   as the reference option for the BIER-TE control plane, but other
   options are equally feasible.  The BIER-TE control plane could
   equally be implemented without automated configuration/protocols, by
   an operator via a CLI on the BFRs.  In that case, operator-configured
   local policy on the BFIR would have to determine how to set the
   appropriate BIER header fields.  The BIER-TE control plane could also
   be decentralized and/or distributed, but this document does not
   consider any additional protocols and/or procedures that would then
   be necessary to coordinate its (distributed/decentralized) entities
   to achieve the above-described functionality.

3.2.1.1.  BIER-TE Topology Discovery and Creation



   The first item listed for BIER-TE topology control (Section 3.2,
   point 1.a.) includes network topology discovery and BIER-TE topology
   creation.  The latter describes the process by which a controller
   determines which routers are to be configured as BFRs and the
   adjacencies between them.

   In statically managed networks, e.g., industrial environments, both
   discovery and creation can be a manual/offline process.

   In other networks, topology discovery may rely on such protocols as
   those that include extending an IGP based on a link-state protocol
   into the BIER-TE controller itself, e.g., BGP-LS [RFC7752] or YANG
   topology [RFC8345], as well as methods specific to BIER-TE -- for
   example, via [BIER-TE-YANG].  These options are non-exhaustive.

   Dynamic creation of the BIER-TE topology can be as easy as mapping
   the network topology 1:1 to the BIER-TE topology by assigning a BP
   for every network subnet adjacency.  In larger networks, it likely
   involves more complex policy and optimization decisions, including
   how to minimize the number of BPs required and how to assign BPs
   across different BitStrings to minimize the number of duplicate
   packets across links when delivering an overlay flow to BFERs using
   different SIs:BitStrings.  These topics are discussed in Section 5.

   When the BIER-TE topology has been determined, the BIER-TE controller
   pushes the BPs/adjacencies to the BIFT of the BFRs.  On each BFR,
   only those SIs:BPs that are adjacencies to other BFRs in the BIER-TE
   topology are populated.

   Communications between the BIER-TE controller and BFRs for both BIER-
   TE topology control and BIER-TE tree control are ideally via
   standardized protocols and data models such as NETCONF/RESTCONF/YANG/
   PCEP.  A vendor-specific CLI on the BFRs is also an option (as in
   many other "Software-Defined Network" (SDN) solutions lacking
   definitions of standardized data models).

3.2.1.2.  Engineered Trees via BitStrings



   In BIER, the same set of BFERs in a single subdomain is always
   encoded as the same BitString.  In BIER-TE, the BitString used to
   reach the same set of BFERs in the same subdomain can be different
   for different overlay flows because the BitString encodes the paths
   towards the BFERs, so the BitStrings from different BFIRs to the same
   set of BFERs will often be different.  Likewise, the BitString from
   the same BFIR to the same set of BFERs can be different for different
   overlay flows if different policies should be applied to those
   overlay flows, such as shortest path trees, Steiner trees (minimum
   cost trees), diverse path trees for redundancy, and so on.

   See also [BIER-MCAST-OVERLAY] for an application leveraging BIER-TE
   engineered trees.

3.2.1.3.  Changes in the Network Topology



   If the network topology changes (not failure based) so that
   adjacencies that are assigned to bit positions are no longer needed,
   the BIER-TE controller can reuse those bit positions for new
   adjacencies.  First, these bit positions need to be removed from any
   BFIR flow state and BFR BIFT state.  Then, they can be repopulated,
   first into the BIFT and then into the BFIR.

3.2.1.4.  Link/Node Failures and Recovery



   When links or nodes fail or recover in the topology, BIER-TE could
   quickly respond with "Fast Reroute" (FRR) procedures such as those
   described in [BIER-TE-PROTECTION], the details of which are out of
   scope for this document.  It can also more slowly react by
   recalculating the BitStrings of affected multicast flows.  This
   reaction is slower than the FRR procedure because the BIER-TE
   controller needs to receive link/node up/down indications,
   recalculate the desired BitStrings, and push them down into the
   BFIRs.  With FRR, this is all performed locally on a BFR receiving
   the adjacency up/down notification.

3.3.  The BIER-TE Forwarding Plane



   The BIER-TE forwarding plane consists of the following components:

   1.  On a BFIR, imposition of the BIER header for packets from overlay
       flows.  This is driven by state established by the BIER-TE
       control plane, the multicast flow overlay as explained in
       Section 3.1, or a combination of both.

   2.  On BFRs (including BFIRs and BFERs), forwarding/replication of
       BIER packets according to their SD, SI, "BitStringLength" (BSL),
       BitString, and, optionally, entropy fields as explained in
       Section 4.  Processing of other BIER header fields, such as the
       "Differentiated Services Code Point" (DSCP) field, is outside the
       scope of this document.

   3.  On BFERs, removal of the BIER header and dispatching of the
       payload according to state created by the BIER-TE control plane
       and/or overlay layer.

   When the BIER-TE forwarding plane receives a packet, it simply looks
   up the bit positions that are set in the BitString of the packet in
   the BIFT that was populated by the BIER-TE controller.  For every BP
   that is set in the BitString and has one or more adjacencies in the
   BIFT, a copy is made according to the types of adjacencies for that
   BP in the BIFT.  Before sending any copies, the BFR clears all BPs in
   the BitString of the packet for which the BFR has one or more
   adjacencies in the BIFT.  Clearing these bits prevents packets from
   looping when a BitString erroneously includes a forwarding loop.
   When a forward_connected() adjacency has the "DoNotClear" (DNC) flag
   set, this BP is reset for the packet copied to that adjacency.  See
   Section 4.2.1.

3.4.  The Routing Underlay



   For forward_connected() adjacencies, BIER-TE sends BIER packets to
   directly connected BIER-TE neighbors as L2 (unicast) BIER packets
   without requiring a routing underlay.  For forward_routed()
   adjacencies, BIER-TE forwarding encapsulates a copy of the BIER
   packet so that it can be delivered by the forwarding plane of the
   routing underlay to the routable destination address indicated in the
   adjacency.  See Section 4.2.2 for details on forward_routed()
   adjacencies.

   BIER relies on the routing underlay to calculate paths towards BFERs
   and derive next-hop BFR adjacencies for those paths.  These two steps
   commonly rely on BIER-specific extensions to the routing protocols of
   the routing underlay but may also be established by a controller.  In
   BIER-TE, the next hops for a packet are determined by the BitString
   through the BIER-TE controller-established adjacencies on the BFR for
   the BPs of the BitString.  There is thus no need for BFR-specific
   routing underlay extensions to forward BIER packets with BIER-TE
   semantics.

   Encapsulation parameters can be provisioned by the BIER-TE controller
   into the forward_connected() or forward_routed() adjacencies directly
   without relying on a routing underlay.

   If the BFR intends to support FRR for BIER-TE, then the BIER-TE
   forwarding plane needs to receive fast adjacency up/down
   notifications: link up/down or neighbor up/down, e.g., from
   "Bidirectional Forwarding Detection" (BFD).  Providing these
   notifications is considered to be part of the routing underlay in
   this document.

3.5.  Traffic Engineering Considerations



   Traffic Engineering [TE-OVERVIEW] provides performance optimization
   of operational IP networks while utilizing network resources
   economically and reliably.  The key elements needed to effect Traffic
   Engineering are policy, path steering, and resource management.
   These elements require support at the control/controller level and
   within the forwarding plane.

   Policy decisions are made within the BIER-TE control plane, i.e.,
   within BIER-TE controllers.  Controllers use policy when composing
   BitStrings and BFR BIFT state.  The mapping of user/IP traffic to
   specific BitStrings / BIER-TE flows is made based on policy.  The
   specific details of BIER-TE policies and how a controller uses them
   are out of scope for this document.

   Path steering is supported via the definition of a BitString.
   BitStrings used in BIER-TE are composed based on policy and resource
   management considerations.  For example, when composing BIER-TE
   BitStrings, a controller must take into account the resources
   available at each BFR and for each BP when it is providing
   congestion-loss-free services such as Rate-Controlled Service
   Disciplines [RCSD94].  Resource availability could be provided, for
   example, via routing protocol information but may also be obtained
   via a BIER-TE control protocol such as NETCONF or any other protocol
   commonly used by a controller to understand the resources of the
   network on which it operates.  The resource usage of the BIER-TE
   traffic admitted by the BIER-TE controller can be solely tracked on
   the BIER-TE controller based on local accounting as long as no
   forward_routed() adjacencies are used (see Section 4.2.2 for the
   definition of forward_routed() adjacencies).  When forward_routed()
   adjacencies are used, the paths selected by the underlying routing
   protocol need to be tracked as well.

   Resource management has implications for the forwarding plane beyond
   the BIER-TE-defined steering of packets; this includes allocation of
   buffers to guarantee the worst-case requirements for admitted RCSD
   traffic and potentially policing and/or rate-shaping mechanisms,
   typically done via various forms of queuing.  This level of resource
   control, while optional, is important in networks that wish to
   support congestion management policies to control or regulate the
   offered traffic to deliver different levels of service and alleviate
   congestion problems, or those networks that wish to control latencies
   experienced by specific traffic flows.

4.  BIER-TE Forwarding

4.1.  The BIER-TE Bit Index Forwarding Table (BIFT)



   The BIER-TE BIFT is equivalent to the (non-TE) BIER BIFT.  It exists
   on every BFR running BIER-TE.  For every BIER "subdomain" (SD) in use
   for BIER-TE, the BIFT is constructed per the example shown in
   Figure 4.  The BIFT in the figure assumes a BSL of 8 "bit positions"
   (BPs) in the packets BitString.  As in [RFC8279], this BSL is purely
   used as an example and is not a BSL supported by BIER/BIER-TE
   (minimum BSL is 64).

   A BIER-TE BIFT is compared to a BIER BIFT as shown in [RFC8279] as
   follows.

   In both BIER and BIER-TE, BIFT rows/entries are indexed in their
   respective BIER pseudocode ([RFC8279], Section 6.5) and BIER-TE
   pseudocode (4.4">Section 4.4) by the BIFT-index derived from the packet's
   SI, BSL, and the one bit position of the packets BitString (BP)
   addressing the BIFT row: BIFT-index = SI * BSL + BP - 1.  BPs within
   a BitString are numbered from 1 to BSL -- hence, the - 1 offset when
   converting to a BIFT-index.  This document also uses the notion
   "SI:BP" to indicate BIFT rows.  [RFC8279] uses the equivalent notion
   "SI:BitString", where the BitString is filled with only the BPs for
   the BIFT row.

   In BIER, each BIFT-index addresses one BFER by its BFR-id = BIFT-
   index + 1 and is populated on each BFR with the next-hop "BFR
   Neighbor" (BFR-NBR) towards that BFER.

   In BIER-TE, each BIFT-index and, therefore, SI:BP indicates one or,
   in the case of reuse of SI:BP, more than one adjacency between BFRs
   in the topology.  The SI:BP is populated with the adjacency on the
   upstream BFR of the adjacency.  The BIFT entries are empty on all
   other BFRs.

   In BIER, each BIFT row also requires a "Forwarding Bit Mask" (F-BM)
   entry for BIER forwarding rules.  In BIER-TE forwarding, an F-BM is
   not required but can be used when implementing BIER-TE on forwarding
   hardware, derived from BIER forwarding, that must use an F-BM.  This
   is discussed in the first variation of BIER-TE forwarding pseudocode
   shown in Section 4.4.

    -------------------------------------------------------------------
    | BIFT-index |      | Adjacencies:                                |
    | (SI:BP)    |(F-BM)| <empty> or one or more per entry            |
    ===================================================================
    |               BIFT indices for Packets with SI=0                |
    -------------------------------------------------------------------
    | 0 (0:1)    | ...  | forward_connected(interface,neighbor{,DNC}) |
    -------------------------------------------------------------------
    | 1 (0:2)    | ...  | forward_connected(interface,neighbor{,DNC}) |
    |            | ...  | forward_connected(interface,neighbor{,DNC}) |
    -------------------------------------------------------------------
    |  ...       | ...  | ...                                         |
    -------------------------------------------------------------------
    | 4 (0:5)    | ...  | local_decap({VRF})                          |
    -------------------------------------------------------------------
    | 5 (0:6)    | ...  | forward_routed({VRF,}l3-neighbor)           |
    -------------------------------------------------------------------
    | 6 (0:7)    | ...  | <empty>                                     |
    -------------------------------------------------------------------
    | 7 (0:8)    | ...  | ECMP((adjacency1,...adjacencyN){,seed})     |
    -------------------------------------------------------------------
    |           BIFT indices for BitString/Packet with SI=1           |
    -------------------------------------------------------------------
    | 9 (1:1)    |      | ...                                         |
    |  ...       | ...  | ...                                         |
    -------------------------------------------------------------------

          Figure 4: BIER-TE Bit Index Forwarding Table (BIFT) with
                           Different Adjacencies

   The BIFT is configured for the BIER-TE data plane of a BFR by the
   BIER-TE controller through an appropriate protocol and data model.
   The BIFT is then used to forward packets, according to the procedures
   for the BIER-TE forwarding plane as specified in Section 3.3.

   Note that a BIFT-index (SI:BP) may be populated in the BIFT of more
   than one BFR to save BPs.  See Section 5.1.6 for an example of how a
   BIER-TE controller could assign BPs to (logical) adjacencies shared
   across multiple BFRs, Section 5.1.3 for an example of assigning the
   same BP to different adjacencies, and Section 5.1.9 for general
   guidelines regarding the reuse of BPs across different adjacencies.

   {VRF} indicates the Virtual Routing and Forwarding context into which
   the BIER payload is to be delivered.  This is optional and depends on
   the multicast flow overlay.

4.2.  Adjacency Types



4.2.1.  Forward Connected



   A "forward_connected()" adjacency is an adjacency towards a directly
   connected BFR-NBR using an interface address of that BFR on the
   connecting interface.  A forward_connected() adjacency does not route
   packets; only L2 forwards them to the neighbor.

   Packets sent to an adjacency with "DoNotClear" (DNC) set in the BIFT
   MUST NOT have the bit position for that adjacency cleared when the
   BFR creates a copy for it.  The bit position will still be cleared
   for copies of a packet made towards other adjacencies.  This can be
   used, for example, in ring topologies as explained in Section 5.1.6.

   For protection against loops caused by misconfiguration (see
   Section 5.2.1), DNC is only permissible for forward_connected()
   adjacencies.  No need or benefit of DNC for other types of
   adjacencies was identified, and associated risks were not analyzed.

4.2.2.  Forward Routed



   A "forward_routed()" adjacency is an adjacency towards a BFR that
   uses a (tunneling) encapsulation that will cause a packet to be
   forwarded by the routing underlay towards the adjacent BFR indicated
   via the l3-neighbor parameter of the forward_routed() adjacency.
   This can leverage any feasible encapsulation, such as MPLS or
   tunneling over IP/IPv6, as long as the BIER-TE packet can be
   identified as a payload.  This identification can rely on either the
   BIER/BIER-TE co-existence mechanisms described in Section 4.3 or
   explicit support for a BIER-TE payload type in the tunneling
   encapsulation.

   Forward_routed() adjacencies are necessary to pass BIER-TE traffic
   across routers that are not BIER-TE capable or to minimize the number
   of required BPs by tunneling over (BIER-TE-capable) routers on which
   neither replication nor path steering is desired, or simply to
   leverage the routing underlay's path redundancy and FRR towards the
   next BFR.  They may also be useful to a multi-subnet adjacent BFR for
   leveraging the routing underlay ECMP independently of BIER-TE ECMP
   (Section 4.2.3).

4.2.3.  ECMP



   (Non-TE) BIER ECMP is tied to the BIER BIFT processing semantic and
   is therefore not directly usable with BIER-TE.

   A BIER-TE "Equal-Cost Multipath" (ECMP()) adjacency as shown in
   Figure 4 for BIFT-index 7 has a list of two or more non-ECMP()
   adjacencies as parameters and an optional seed parameter.  When a
   BIER-TE packet is copied onto such an ECMP() adjacency, an
   implementation-specific so-called hash function will select one out
   of the list's adjacencies to which the packet is forwarded.  If the
   packet's encapsulation contains an entropy field, the entropy field
   SHOULD be respected; two packets with the same value of the entropy
   field SHOULD be sent on the same adjacency.  The seed parameter
   permits the design of hash functions that are easy to implement at
   high speed without running into polarization issues across multiple
   consecutive ECMP hops.  See Section 5.1.7 for details.

4.2.4.  Local Decap(sulation)



   A "local_decap()" adjacency passes a copy of the payload of the BIER-
   TE packet to the protocol ("NextProto") within the BFR (IP/IPv6,
   Ethernet,...) responsible for that payload according to the packet
   header fields.  A local_decap() adjacency turns the BFR into a BFER
   for matching packets.  Local_decap() adjacencies require the BFER to
   support routing or switching for NextProto to determine how to
   further process the packets.

4.3.  Encapsulation / Co-existence with BIER



   Specifications for BIER-TE encapsulation are outside the scope of
   this document.  This section gives explanations and guidelines.

   The handling of "Maximum Transmission Unit" (MTU) limitations is
   outside the scope of this document and is not discussed in [RFC8279]
   either.  Instead, this process is part of the BIER-TE packet
   encapsulation and/or flow overlay; for example, see [RFC8296],
   Section 3.  It applies equally to BIER-TE and BIER.

   Because a BFR needs to interpret the BitString of a BIER-TE packet
   differently from a (non-TE) BIER packet, it is necessary to
   distinguish BIER packets from BIER-TE packets.  In BIER encapsulation
   [RFC8296], the BIFT-id field of the packet indicates the BIFT of the
   packet.  BIER and BIER-TE can therefore be run simultaneously, when
   the BIFT-id address space is shared across BIER BIFTs and BIER-TE
   BIFTs.  Partitioning the BIFT-id address space is subject to BIER-TE/
   BIER control plane procedures.

   When [RFC8296] is used for BIER with MPLS, BIFT-id address ranges can
   be dynamically allocated from MPLS label space only for the set of
   actually used SD:BSL BIFTs.  This also permits the allocation of non-
   overlapping label ranges for BIFT-ids that are to be used with BIER-
   TE BIFTs.

   With MPLS, it is also possible to reuse the same SD space for both
   BIER-TE and BIER, so that the same SD has both a BIER BIFT with a
   corresponding range of BIFT-ids and disjoint BIER-TE BIFTs with a
   non-overlapping range of BIFT-ids.

   Assume that a fixed mapping from BSL, SD, and SI to a BIFT-id is
   used, which does not explicitly partition the BIFT-id space between
   BIER and BIER-TE -- for example, as proposed for non-MPLS forwarding
   with BIER encapsulation [RFC8296] in [NON-MPLS-BIER-ENCODING],
   Section 5.  In this case, it is necessary to allocate disjoint SDs to
   BIER and BIER-TE BIFTs so that both can be addressed by the BIFT-ids.
   The encoding proposed in Section 6 of [NON-MPLS-BIER-ENCODING] does
   not statically encode the BSL or SD into the BIFT-id, but the
   encoding permits a mapping and hence could provide the same freedom
   as when MPLS is being used (the same SD, or different SDs for BIER/
   BIER-TE).

   Forward_routed() requires an encapsulation that permits directing
   unicast encapsulated BIER-TE packets to a specific interface address
   on a target BFR.  With MPLS encapsulation, this can simply be done
   via a label stack with that address's label as the top label,
   followed by the label assigned to the (BSL,SD,SI) BitString.  With
   non-MPLS encapsulation, some form of IP encapsulation would be
   required (for example, IP/GRE).

   The encapsulation used for forward_routed() adjacencies can equally
   support existing advanced adjacency information such as "loose source
   routes" via, for example, MPLS label stacks or appropriate header
   extensions (e.g., for IPv6).

4.4.  BIER-TE Forwarding Pseudocode



   The pseudocode for BIER-TE forwarding, as shown in Figure 5, is based
   on the (non-TE) BIER forwarding pseudocode provided in [RFC8279],
   Section 6.5, with one modification.

      void ForwardBitMaskPacket_withTE (Packet)
      {
          SI=GetPacketSI(Packet);
          Offset=SI*BitStringLength;
          for (Index = GetFirstBitPosition(Packet->BitString); Index ;
               Index = GetNextBitPosition(Packet->BitString, Index)) {
              F-BM = BIFT[Index+Offset]->F-BM;
              if (!F-BM) continue;                            [3]
              BFR-NBR = BIFT[Index+Offset]->BFR-NBR;
              PacketCopy = Copy(Packet);
              PacketCopy->BitString &= F-BM;                  [2]
              PacketSend(PacketCopy, BFR-NBR);
              // The following must not be done for BIER-TE:
              // Packet->BitString &= ~F-BM;                  [1]
          }
      }

      Figure 5: BIER-TE Forwarding Pseudocode for Required Functions,
                          Based on BIER Pseudocode

   In step [2], the F-BM is used to clear one or more bits in
   PacketCopy.  This step exists in both BIER and BIER-TE, but the F-BMs
   need to be populated differently for BIER-TE than for BIER for the
   desired clearing.

   In BIER, multiple bits of a BitString can have the same BFR-NBR.
   When a received packets BitString has more than one of those bits
   set, BIER's replication logic has to prevent more than one PacketCopy
   from being sent to that BFR-NBR ([1]).  Likewise, the PacketCopy sent
   to a BFR-NBR must clear all bits in its BitString that are not routed
   across a BFR-NBR.  This prevents BIER's replication logic from
   creating duplicates on any possible further BFRs ([2]).

   To solve both [1] and [2] for BIER, the F-BM of each bit index needs
   to have all bits set that this BFR wants to route across a BFR-
   NBR.  [2] clears all other bits in PacketCopy->BitString, and [1]
   clears those bits from Packet->BitString after the first PacketCopy.

   In BIER-TE, a BFR-NBR in this pseudocode is an adjacency --
   forward_connected(), forward_routed(), or local_decap().  There is no
   need for [2] to suppress duplicates in the same way that BIER does,
   because in general, different BPs would never have the same
   adjacency.  If a BIER-TE controller actually finds some optimization
   in which this would be desirable, then the controller is also
   responsible for ensuring that only one of those bits is set in any
   Packet->BitString, unless the controller explicitly wants duplicates
   to be created.

   The following points describe how the F-BM for each BP is configured
   in the BIFT and how this impacts the BitString of the packet being
   processed with that BIFT:

   1.  The F-BMs of all BIFT BPs without an adjacency have all their
       bits clear.  This will cause [3] to skip further processing of
       such a BP.

   2.  All BIFT BPs with an adjacency (with the DNC flag clear) have an
       F-BM that has only those BPs set for which this BFR does not have
       an adjacency.  This causes [2] to clear all bits from
       PacketCopy->BitString for which this BFR does have an adjacency.

   3.  [1] is not performed for BIER-TE.  All bit clearing required by
       BIER-TE is performed by [2].

   This forwarding pseudocode can support the required BIER-TE
   forwarding functions (see Section 4.5) -- forward_connected(),
   forward_routed(), and local_decap() -- but cannot support the
   recommended functions (DNC flag and multiple adjacencies per bit) or
   the optional function (i.e., ECMP() adjacencies).  The DNC flag
   cannot be supported when using only [1] to mask bits.

   The modified and expanded forwarding pseudocode in Figure 6 specifies
   how to support all BIER-TE forwarding functions (required,
   recommended, and optional):

   1.  This pseudocode eliminates per-bit F-BMs, therefore reducing the
       size of BIFT state by SI*BSL^2 and eliminating the need for per-
       packet-copy BitString masking operations, except for adjacencies
       with the DNC flag set:

       1.a  AdjacentBits[SI] are bit positions with a non-empty list of
            adjacencies in this BFR BIFT.  This can be computed whenever
            the BIER-TE controller updates (adds/removes) adjacencies in
            the BIFT.

       1.b  The BFR needs to create packet copies for these adjacent
            bits when they are set in the packets BitString.  This set
            of bits is calculated in PktAdjacentBits.

       1.c  All bit positions for which the BFR creates copies have to
            be cleared in packet copies to avoid loops.  This is done by
            masking the BitString of the packet with ~AdjacentBits[SI].
            When an adjacency has DNC set, this bit position is set
            again only for the packet copy towards that bit position.

   2.  BIFT entries may contain more than one adjacency in support of
       specific configurations, such as a hub and multiple spokes
       (Section 5.1.5).  The code therefore includes a loop over these
       adjacencies.

   3.  The ECMP() adjacency is also shown in the figure.  Its parameters
       are a seed and "ListOfAdjacencies", from which one is picked.

   4.  The forward_connected(), forward_routed(), and local_decap()
       adjacencies are shown with their parameters.

    void ForwardBitMaskPacket_withTE (Packet)
    {
        SI = GetPacketSI(Packet);
        Offset = SI * BitStringLength;
        // Determine adjacent bits in the packets BitString
        PktAdjacentBits = Packet->BitString & AdjacentBits[SI];

        // Clear adjacent bits in the packet header to avoid loops
        Packet->BitString &= ~AdjacentBits[SI];

        // Loop over PktAdjacentBits to create packet copies
        for (Index = GetFirstBitPosition(PktAdjacentBits); Index ;
             Index = GetNextBitPosition(PktAdjacentBits, Index)) {
            for adjacency in BIFT[Index+Offset]->Adjacencies {
                if(adjacency.type == ECMP(ListOfAdjacencies,seed) ) {
                    I = ECMP_hash(sizeof(ListOfAdjacencies),
                                  Packet->Entropy,seed);
                    adjacency = ListOfAdjacencies[I];
                }
                PacketCopy = Copy(Packet);
                switch(adjacency.type) {
                    case forward_connected(interface,neighbor,DNC):
                        if(DNC)
                            PacketCopy->BitString |= 1<<(Index-1);
                        SendToL2Unicast(PacketCopy,interface,neighbor);

                    case forward_routed({VRF,}l3-neighbor):
                        SendToL3(PacketCopy,{VRF,}l3-neighbor);

                    case local_decap({VRF},neighbor):
                        DecapBierHeader(PacketCopy);
                        PassTo(PacketCopy,{VRF,}Packet->NextProto);
                }
            }
        }
    }

       Figure 6: Complete BIER-TE Forwarding Pseudocode for Required,
                    Recommended, and Optional Functions

4.5.  BFR Requirements for BIER-TE Forwarding



   BFRs that support BIER-TE and BIER MUST support a configuration that
   enables BIER-TE instead of (non-TE) BIER forwarding rules for all
   BIFTs of one or more BIER subdomains.  Every BP in a BIER-TE BIFT
   MUST support having zero or one adjacency.  BIER-TE forwarding MUST
   support the adjacency types forward_connected() with the DNC flag not
   set, forward_routed(), and local_decap().  As explained in
   Section 4.4, these required BIER-TE forwarding functions can be
   implemented via the same forwarding pseudocode as that used for BIER
   forwarding, except for one modification (skipping one masking with an
   F-BM).

   BIER-TE forwarding SHOULD support forward_connected() adjacencies
   with the DNC flag set, as this is very useful for saving bits in
   rings (see Section 5.1.6).

   BIER-TE forwarding SHOULD support more than one adjacency on a bit.
   This allows bits to be saved in hub-and-spoke scenarios (see
   Section 5.1.5).

   BIER-TE forwarding MAY support ECMP() adjacencies to save bits in
   ECMP scenarios; see Section 5.1.7 for an example.  This is an
   optional requirement, because for ECMP deployments using BIER-TE one
   can also leverage the routing underlay ECMP via forward_routed()
   adjacencies and/or might prefer to have more explicit control of the
   path chosen via explicit BPs/adjacencies for each ECMP path
   alternative.

5.  BIER-TE Controller Operational Considerations



5.1.  Bit Position Assignments



   This section describes how the BIER-TE controller can use the
   different BIER-TE adjacency types to define the bit positions of a
   BIER-TE domain.

   Because the size of the BitString limits the size of the BIER-TE
   domain, many of the options described here exist to support larger
   topologies with fewer bit positions.

5.1.1.  P2P Links



   On a "point-to-point" (P2P) link that connects two BFRs, the same bit
   position can be used on both BFRs for the adjacency to the
   neighboring BFR.  A P2P link therefore requires only one bit
   position.

5.1.2.  BFERs



   Every non-leaf BFER is given a unique bit position with a
   local_decap() adjacency.

5.1.3.  Leaf BFERs



   A leaf BFER is one where incoming BIER-TE packets never need to be
   forwarded to another BFR but are only sent to the BFER to exit the
   BIER-TE domain.  For example, in networks where "Provider Edge" (PE)
   routers are spokes connected to Provider (P) routers, those PEs are
   leaf BFERs, unless there is a U-turn between two PEs.

   Consider how redundant disjoint traffic can reach BFER1/BFER2 as
   shown in Figure 7: when BFER1/BFER2 are non-leaf BFERs as shown on
   the right-hand side, one traffic copy would be forwarded to BFER1
   from BFR1, but the other one could only reach BFER1 via BFER2, which
   makes BFER2 a non-leaf BFER.  Likewise, BFER1 is a non-leaf BFER when
   forwarding traffic to BFER2.  Note that the BFERs on the left-hand
   side of the figure are only guaranteed to be leaf BFERs by correctly
   applying a routing configuration that prohibits transit traffic from
   passing through the BFERs, which is commonly applied in these
   topologies.

           BFR1(P) BFR2(P)             BFR1(P)  BFR2(P)
             |  \ /  |                    |       |
             |   X   |                    |       |
             |  / \  |                    |       |
        BFER1(PE)  BFER2(PE)        BFER1(PE)----BFER2(PE)

                                              ^ U-turn link

            Leaf BFER /               Non-leaf BFER /
             PE router                  PE router

                  Figure 7: Leaf vs. Non-Leaf BFER Example

   In most situations, leaf BFERs that are to be addressed via the same
   BitString can share a single bit position for their local_decap()
   adjacency in that BitString and therefore save bit positions.  On a
   non-leaf BFER, a received BIER-TE packet may only need to transit the
   BFER, or it may also need to be decapsulated.  Whether or not to
   decapsulate the packet therefore needs to be indicated by a unique
   bit position populated only on the BIFT of this BFER with a
   local_decap() adjacency.  On a leaf BFER, packets never need to pass
   through; any packet received is therefore usually intended to be
   decapsulated.  This can be expressed by a single, shared bit position
   that is populated with a local_decap() adjacency on all leaf BFERs
   addressed by the BitString.

   The possible exceptions to this leaf BFER bit position optimization
   scenario can be cases where the bit position on the prior BIER-TE BFR
   (which created the packet copy for the leaf BFER in question) is
   populated with multiple adjacencies as an optimization -- for
   example, as described in Sections 5.1.4 and 5.1.5.  With either of
   these two optimizations, the sender of the packet could only control
   explicitly whether the packet was to be decapsulated on the leaf BFER
   in question, if the leaf BFER has a unique bit position for its
   local_decap() adjacency.

   However, if the bit position is shared across a leaf BFER and packets
   are therefore decapsulated -- potentially unnecessarily -- this may
   still be appropriate if the decapsulated payload of the BIER-TE
   packet indicates whether or not the packets need to be further
   processed/received.  This is typically true, for example, if the
   payload is IP multicast, because IP multicast on a BFER would know
   the membership state of the IP multicast payload and be able to
   discard it if the packets were delivered unnecessarily by the BIER-TE
   layer.  If the payload has no such membership indication and the BFIR
   wants to have explicit control regarding which BFERs are to receive
   and decapsulate a packet, then these two optimizations cannot be used
   together with shared bit position optimization for a leaf BFER.

5.1.4.  LANs



   In a LAN, the adjacency to each neighboring BFR is given a unique bit
   position.  The adjacency of this bit position is a
   forward_connected() adjacency towards the BFR, and this bit position
   is populated into the BIFT of all the other BFRs on that LAN.

                                    BFR1
                                     |p1
                              LAN1-+-+---+-----+
                                 p3|   p4|   p2|
                                 BFR3  BFR4  BFR7

                           Figure 8: LAN Example

   If bandwidth on the LAN is not an issue and most BIER-TE traffic
   should be copied to all neighbors on a LAN, then bit positions can be
   saved by assigning just a single bit position to the LAN and
   populating the bit position of the BIFTs of each BFR on the LAN with
   a list of forward_connected() adjacencies to all other neighbors on
   the LAN.

   This optimization does not work in the case of BFRs redundantly
   connected to more than one LAN with this optimization.  These BFRs
   would receive duplicates and forward those duplicates into the other
   LANs.  Such BFRs require separate bit positions for each LAN they
   connect to.

5.1.5.  Hub and Spoke



   In a setup with a hub and multiple spokes connected via separate P2P
   links to the hub, all P2P adjacencies from the hub to the spokes'
   links can share the same bit position.  The bit position on the hub's
   BIFT is set up with a list of forward_connected() adjacencies, one
   for each spoke.

   This option is similar to the bit position optimization in LANs:
   redundantly connected spokes need their own bit positions, unless
   they are themselves leaf BFERs.

   This type of optimized BP could be used, for example, when all
   traffic is "broadcast" traffic (very dense receiver sets), such as
   live TV or many-to-many telemetry, including situational awareness.
   This BP optimization can then be used to explicitly steer different
   traffic flows across different ECMP paths in data-center or
   broadband-aggregation networks with minimal use of BPs.

5.1.6.  Rings



   In L3 rings, instead of assigning a single bit position for every P2P
   link in the ring, it is possible to save bit positions by setting the
   "DoNotClear" (DNC) flag on forward_connected() adjacencies.

   For the ring shown in Figure 9, a single bit position will suffice to
   forward traffic entering the ring at BFRa or BFRb all the way up to
   BFR1, as follows.

   On BFRa, BFRb, BFR30,... BFR3, the bit position is populated with a
   forward_connected() adjacency pointing to the clockwise neighbor on
   the ring and with DNC set.  On BFR2, the adjacency also points to the
   clockwise neighbor BFR1, but without DNC set.

   Handling DNC this way ensures that copies forwarded from any BFRs in
   the ring to a BFR outside the ring will not have the ring bit
   position set, therefore minimizing the risk of creating loops.

                  v        v
                  |        |
           L1     |   L2   |   L3
       /-------- BFRa ---- BFRb --------------------\
       |                                            |
       \- BFR1 - BFR2 - BFR3 - ... - BFR29 - BFR30 -/
           |      |    L4               |      |
        p33|                         p15|
           BFRd                       BFRc

                           Figure 9: Ring Example

   Note that this example only permits packets intended to make it all
   the way around the ring to enter it at BFRa and BFRb.  Note also that
   packets will always travel clockwise.  If packets should be allowed
   to enter the ring at any of the ring's BFRs, then one would have to
   use two ring bit positions, one for each direction: clockwise and
   counterclockwise.

   Both would be set up to stop rotating on the same link, e.g., L1.
   When the ring's BFIR creates the clockwise copy, it will clear the
   counterclockwise bit position because the DNC bit only applies to the
   bit for which the replication is done (likewise for the clockwise bit
   position for the counterclockwise copy).  As a result, the ring's
   BFIR will send a copy in both directions, serving BFRs on either side
   of the ring up to L1.

5.1.7.  Equal-Cost Multipath (ECMP)



   An ECMP() adjacency allows the use of just one BP to deliver packets
   to one of N adjacencies instead of one BP for each adjacency.  In the
   common example case shown in Figure 10, a link bundle of three links
   L1,L2,L3 connects BFR1 and BFR2, and only one BP is used instead of
   three BPs to deliver packets from BFR1 to BFR2.

                --L1-----
           BFR1 --L2----- BFR2
                --L3-----

     BIFT entry in BFR1:
     ------------------------------------------------------------------
     | Index |  Adjacencies                                           |
     ==================================================================
     | 0:6   |  ECMP({forward_connected(L1, BFR2),                    |
     |       |        forward_connected(L2, BFR2),                    |
     |       |        forward_connected(L3, BFR2)}, seed)             |
     ------------------------------------------------------------------

     BIFT entry in BFR2:
     ------------------------------------------------------------------
     | Index |  Adjacencies                                           |
     ==================================================================
     | 0:6   |  ECMP({forward_connected(L1, BFR1),                    |
     |       |        forward_connected(L2, BFR1),                    |
     |       |        forward_connected(L3, BFR1)}, seed)             |
     ------------------------------------------------------------------

                          Figure 10: ECMP Example

   This document does not standardize any ECMP algorithm because it is
   sufficient for implementations to document their freely chosen ECMP
   algorithm.  Figure 11 shows an example ECMP algorithm and would
   double as its documentation: a BIER-TE controller could determine
   which adjacency is chosen based on the seed and adjacencies
   parameters and on packet entropy.

      forward(packet, ECMP(adj(0), adj(1),... adj(N-1), seed)):
         i = (packet(bier-header-entropy) XOR seed) % N
         forward packet to adj(i)

                     Figure 11: ECMP Algorithm Example

   In the example shown in Figure 12, all traffic from BFR1 towards
   BFR10 is intended to be ECMP load-split equally across the topology.
   This example is not meant as a likely setup; rather, it illustrates
   that ECMP can be used to share BPs not only across link bundles but
   also across alternative paths across different transit BFRs, and it
   explains the use of the seed parameter.

                    BFR1         (BFIR)
                  /L11  \L12
                 /       \
             BFR2         BFR3
            /L21 \L22    /L31 \L32
           /      \     /      \
          BFR4  BFR5   BFR6  BFR7
           \      /     \      /
            \    /       \    /
             BFR8         BFR9
                 \       /
                  \     /
                   BFR10         (BFER)

     BIFT entry in BFR1:
     ------------------------------------------------------------------
     | 0:6   |  ECMP({forward_connected(L11, BFR2),                   |
     |       |        forward_connected(L12, BFR3)}, seed1)           |
     ------------------------------------------------------------------

     BIFT entry in BFR2:
     ------------------------------------------------------------------
     | 0:7   |  ECMP({forward_connected(L21, BFR4),                   |
     |       |        forward_connected(L22, BFR5)}, seed1)           |
     ------------------------------------------------------------------

     BIFT entry in BFR3:
     ------------------------------------------------------------------
     | 0:7   |  ECMP({forward_connected(L31, BFR6),                   |
     |       |        forward_connected(L32, BFR7)}, seed1)           |
     ------------------------------------------------------------------

     BIFT entry in BFR4, BFR5:
     ------------------------------------------------------------------
     | 0:8   |  forward_connected(Lxx, BFR8)  |xx differs on BFR4/BFR5|
     ------------------------------------------------------------------

     BIFT entry in BFR6, BFR7:
     ------------------------------------------------------------------
     | 0:8   |  forward_connected(Lxx, BFR9)  |xx differs on BFR6/BFR7|
     ------------------------------------------------------------------

     BIFT entry in BFR8, BFR9:
     ------------------------------------------------------------------
     | 0:9   |  forward_connected(Lxx, BFR10) |xx differs on BFR8/BFR9|
     ------------------------------------------------------------------

                      Figure 12: Polarization Example

   Note that for the following discussion of ECMP, only the BIFT ECMP()
   adjacencies on BFR1, BFR2, and BFR3 are relevant.  The reuse of BPs
   across BFRs in this example is further explained in Section 5.1.9
   below.

   With the ECMP setup shown in the topology above, traffic would not be
   equally load-split.  Instead, links L22 and L31 would see no traffic
   at all: BFR2 will only see traffic from BFR1, for which the ECMP hash
   in BFR1 selected the first adjacency in the list of two adjacencies
   given as parameters to the ECMP: link L11-to-BFR2.  BFR2 again
   performs ECMP with two adjacencies on that subset of traffic using
   the same seed1 and will therefore again select the first of its two
   adjacencies: L21-to-BFR4.  Therefore, L22 and BFR5 see no traffic
   (likewise for L31 and BFR6).

   This issue in BFR2/BFR3 is called "polarization".  It results from
   the reuse of the same hash function across multiple consecutive hops
   in topologies like these.  To resolve this issue, the ECMP()
   adjacency on BFR1 can be set up with a different seed2 than the
   ECMP() adjacencies on BFR2/BFR3.  BFR2/BFR3 can use the same hash
   because packets will not sequentially pass across both of them.
   Therefore, they can also use the same BP (i.e., 0:7).

   Note that ECMP solutions outside of BIER often hide the seed by auto-
   selecting it from local entropy such as unique local or next-hop
   identifiers.  Allowing the BIER-TE controller to explicitly set the
   seed gives the BIER-TE controller the ability to control the
   selection of the same path or different paths across multiple
   consecutive ECMP hops.

5.1.8.  Forward Routed Adjacencies



5.1.8.1.  Reducing Bit Positions



   Forward_routed() adjacencies can reduce the number of bit positions
   required when the path steering requirement is not hop-by-hop
   explicit path selection but rather is loose-hop selection.
   Forward_routed() adjacencies can also permit BIER-TE operation across
   intermediate-hop routers that do not support BIER-TE.

   Assume that the requirement in Figure 13 is to explicitly steer
   traffic flows that have arrived at BFR1 or BFR4 via a path in the
   routing underlay "Network Area 1" to one of the following next three
   segments: (1) BFR2 via link L1, (2) BFR2 via link L2, or (3) via BFR3
   and then not caring whether the packet is forwarded via L3 or L4.

                      ...............
            ...BFR1--...           ...--L1-- BFR2...
                     ... .Routers. ...--L2--/
            ...BFR4--...           ...--L3-- BFR3...
                     ...           ...--L4--/ |
                      ...............         |
                                             LO
                       Network Area 1

               Figure 13: Forward Routed Adjacencies Example

   To enable this, both BFR1 and BFR4 are set up with a forward_routed()
   adjacency bit position towards an address of BFR2 on link L1, another
   forward_routed() bit position towards an address of BFR2 on link L2,
   and a third forward_routed() bit position towards a node address LO
   of BFR3.

5.1.8.2.  Supporting Nodes without BIER-TE



   Forward_routed() adjacencies also enable incremental deployment of
   BIER-TE.  Only the nodes through which BIER-TE traffic needs to be
   steered -- with or without replication -- need to support BIER-TE.
   Where they are not directly connected to each other, forward_routed()
   adjacencies are used to pass over nodes that are not BIER-TE enabled.

5.1.9.  Reuse of Bit Positions (without DNC)



   BPs can be reused across multiple BFRs to minimize the number of BPs
   needed.  This happens when adjacencies on multiple BFRs use the DNC
   flag as described above, but it can also be done for non-DNC
   adjacencies.  This section only discusses this non-DNC case.

   Because a given BP is cleared when passing a BFR with an adjacency
   for that BP, reusing BPs across multiple BFRs does not introduce any
   problems with duplicates or loops that do not also exist when every
   adjacency has a unique BP.  Instead, the challenge when reusing BPs
   is whether the desired Tree Engineering goals can still be achieved.

   A BP cannot be reused across two BFRs that would need to be passed
   sequentially for some path: the first BFR will clear the BP, so those
   paths cannot be built.  A BP can be set across BFRs that would only
   occur across (A) different paths or (B) different branches of the
   same tree.

   An example of (A) was given in Figure 12, where BP 0:7, BP 0:8, and
   BP 0:9 are each reused across multiple BFRs because a single packet/
   path would never be able to reach more than one BFR sharing the same
   BP.

   Assume that the example was changed: BFR1 has no ECMP() adjacency for
   BP 0:6 but instead has BP 0:5 with forward_connected() to BFR2 and BP
   0:6 with forward_connected() to BFR3.  Packets with both BP 0:5 and
   BP 0:6 would now be able to reach both BFR2 and BFR3, and the still-
   existing reuse of BP 0:7 between BFR2 and BFR3 is a case of (B) where
   reusing a BP is perfect because it does not limit the set of useful
   path choices, as in the following example.

   If instead of reusing BP 0:7 BFR3 used a separate BP 0:10 for its
   ECMP() adjacency, no useful additional path steering options would be
   enabled.  If duplicates at BFR10 were undesirable, this would be done
   by not setting BP 0:5 and BP 0:6 for the same packet.  If the
   duplicates were desirable (e.g., resilient transmission), the
   additional BP 0:10 would also not render additional value.

   Reuse may also save BPs in larger topologies.  Consider the topology
   shown in Figure 14.

                          area1
                      BFR1a BFR1b
                        /    \
           ....................................
           .                Core              .
           ....................................
           |    /       \    /           \  |
         BFR2a BFR2b  BFR3a BFR3b      BFR6a BFR6b
          /-------\   /---------\      /--------\
          | area2 |   |  area3  | ...  | area6  |
          | ring  |   |  ring   |      | ring   |
          \-------/   \---------/      \--------/
          more BFRs    more BFRs        more BFRs

                          Figure 14: Reuse of BPs

   A BFIR/sender (e.g., video headend) is attached to area 1, and the
   five areas 2...6 contain receivers/BFERs.  Assume that each area has
   a distribution ring, each with two BPs to indicate the direction (as
   explained before).  These two BPs could be reused across the five
   areas.  Packets would be replicated through other BPs from the core
   to the desired subset of areas, and once a packet copy reaches the
   ring of the area, the two ring BPs come into play.  This reuse is a
   case of (B), but it limits the topology choices: packets can only
   flow around the same direction in the rings of all areas.  This may
   or may not be acceptable based on the desired path steering options:
   if resilient transmission is the path engineering goal, then it is
   likely a good optimization; however, if the bandwidth of each ring
   were to be optimized separately, it would not be a good limitation.

5.1.10.  Summary of BP Optimizations



   In this section, we reviewed a range of techniques by which a BIER-TE
   controller can create a BIER-TE topology in a way that minimizes the
   number of necessary BPs.

   Without any optimization, a BIER-TE controller would attempt to map
   the network subnet topology 1:1 into the BIER-TE topology, every
   adjacent neighbor in the subnet would require a forward_connected()
   BP, and every BFER would require a local_decap() BP.

   The optimizations described in this document are then as follows:

   1.  P2P links require only one BP (Section 5.1.1).

   2.  All leaf BFERs can share a single local_decap() BP
       (Section 5.1.3).

   3.  A LAN with N BFRs needs at most N BPs (one for each BFR).  It
       only needs one BP for all those BFRs that are not redundantly
       connected to multiple LANs (Section 5.1.4).

   4.  A hub with P2P connections to multiple non-leaf BFER spokes can
       share one BP with all of the spokes if traffic can be flooded to
       all of those spokes, e.g., because of no bandwidth concerns or
       dense receiver sets (Section 5.1.5).

   5.  Rings of BFRs can be built with just two BPs (one for each
       direction), except for BFRs with multiple ring connections --
       similar to LANs (Section 5.1.6).

   6.  ECMP() adjacencies to N neighbors can replace N BPs with one BP.
       Multihop ECMP can avoid polarization through different seeds of
       the ECMP algorithm (Section 5.1.7).

   7.  Forward_routed() adjacencies permit "tunneling" across routers
       that are either BIER-TE capable or not BIER-TE capable where no
       traffic steering or replications are required (Section 5.1.8).

   8.  A BP can generally be reused across a set of nodes where it can
       be guaranteed that no path will ever need to traverse more than
       one node of the set.  Depending on the scenario, this may limit
       the feasible path steering options (Section 5.1.9).

   Note that this list of optimizations is not exhaustive.  Further
   optimizations of BPs are possible, especially when both the set of
   required path steering choices and the possible subsets of BFERs that
   should be able to receive traffic are limited.  The hub-and-spoke
   optimization is a simple example of such traffic-pattern-dependent
   optimizations.

5.2.  Avoiding Duplicates and Loops



5.2.1.  Loops



   Whenever BIER-TE creates a copy of a packet, the BitString of that
   copy will have all bit positions cleared that are associated with
   adjacencies on the BFR.  This prevents packets from looping.  The
   only exceptions are adjacencies with DNC set.

   With DNC set, looping can happen.  Consider in Figure 15 that link L4
   from BFR3 is (inadvertently) plugged into the L1 interface of BFRa
   (instead of BFR2).  This creates a loop where the ring's clockwise
   bit position is never cleared for copies of the packets traveling
   clockwise around the ring.

                  v        v
                  |        |
           L1     |   L2   |   L3
       /-------- BFRa ---- BFRb ---------------------\
       |        .                                    |
       |         ......  Wrong link wiring           |
       |               .                             |
       \- BFR1 - BFR2   BFR3 - ... - BFR29 - BFR30 -/
           |      |    L4               |      |
        p33|                         p15|
           BFRd                       BFRc

                      Figure 15: Miswired Ring Example

   To inhibit looping in the face of such physical misconfiguration,
   only forward_connected() adjacencies are permitted to have DNC set,
   and the link layer port unique unicast destination address of the
   adjacency (e.g., "Media Access Control" (MAC) address) protects
   against closing the loop.  Link layers without port unique link layer
   addresses should not be used with the DNC flag set.

5.2.2.  Duplicates



   Duplicates happen when the graph expressed by a BitString is not a
   tree but is redundantly connecting BFRs with each other.  In
   Figure 16, a BitString of p2,p3,p4,p5 would result in duplicate
   packets arriving on BFER4.  The BIER-TE controller must therefore
   ensure that only BitStrings that are trees are created.

                    BFIR1
                   /    \
                  / p2   \ p3
                 BFR2   BFR3
                  \ p4   / p5
                   \    /
                    BFER4

                       Figure 16: Duplicates Example

   When links are incorrectly physically reconnected before the BIER-TE
   controller updates BitStrings in BFIRs, duplicates can happen.  Like
   loops, these can be inhibited by link layer addressing in
   forward_connected() adjacencies.

   If interface or loopback addresses used in forward_routed()
   adjacencies are moved from one BFR to another, duplicates are equally
   likely to happen.  Such readdressing operations must be coordinated
   with the BIER-TE controller.

5.3.  Managing SIs, Subdomains, and BFR-ids



   When the number of bits required to represent the necessary hops in
   the topology and BFERs exceeds the supported "BitStringLength" (BSL),
   multiple SIs and/or subdomains must be used.  This section discusses
   how this is done.

   BIER-TE forwarding does not require the concept of BFR-ids, but
   routing underlay, flow overlay, and BIER headers may.  This section
   also discusses how BFR-ids can be assigned to BFIRs/BFERs for BIER-
   TE.

5.3.1.  Why SIs and Subdomains?



   For (non-TE) BIER and BIER-TE forwarding, the most important result
   of using multiple SIs and/or subdomains is the same: multicast flow
   overlay packets that need to be sent to BFERs in different SIs or
   subdomains require multiple BIER packets, each one with a BitString
   for a different (SI,subdomain) combination.  Each such BitString uses
   one BSL-sized SI block in the BIFT of the subdomain.  We call this a
   BIFT:SI (block).

   SIs and subdomains have different purposes in the BIER architecture
   and also the BIER-TE architecture.  This impacts how operators manage
   them and especially how flow overlays will likely use them.

   By default, every possible BFIR/BFER in a BIER network would likely
   be given a BFR-id in subdomain 0 (unless there are > 64k BFIRs/
   BFERs).

   If there are different flow services (or service instances) requiring
   replication to different subsets of BFERs, then it will likely not be
   possible to achieve the best replication efficiency for all of these
   service instances via subdomain 0.  Ideal replication efficiency for
   N BFERs exists in a subdomain if they are split over no more than
   ceiling(N/BitStringLength) SIs.

   If service instances justify additional BIER:SI state in the network,
   additional subdomains will be used: BFIRs/BFERs are assigned BFR-ids
   in those subdomains, and each service instance is configured to use
   the most appropriate subdomain.  This results in improved replication
   efficiency for different services.

   Even if creation of subdomains and assignment of BFR-ids to BFIRs/
   BFERs in those subdomains is automated, it is not expected that
   individual service instances can deal with BFERs in different
   subdomains.  A service instance may only support configuration of a
   single subdomain it should rely on.

   To be able to easily reuse (and modify as little as possible)
   existing BIER procedures (including flow overlay and routing
   underlay), when BIER-TE forwarding is added, we therefore reuse SIs
   and subdomains logically in the same way as they are used in BIER:
   all necessary BFIRs/BFERs for a service use a single BIER-TE BIFT and
   are split across as many SIs as necessary (see Section 5.3.2).
   Different services may use different subdomains that primarily exist
   to provide more efficient replication (and, for BIER-TE, desirable
   path steering) for different subsets of BFIRs/BFERs.

5.3.2.  Assigning Bits for the BIER-TE Topology



   In BIER, BitStrings only need to carry bits for BFERs; this leads to
   the model where BFR-ids map 1:1 to each bit in a BitString.

   In BIER-TE, BitStrings need to carry bits to indicate not only the
   receiving BFER but also the intermediate hops/links across which the
   packet must be sent.  The maximum number of BFERs that can be
   supported in a single BitString or BIFT:SI depends on the number of
   bits necessary to represent the desired topology between them.

   "Desired" topology means that it depends on the physical topology and
   the operator's desire to

   1.  permit explicit path steering across every single hop (which
       requires more bits), or

   2.  reduce the number of required bits by exploiting optimizations
       such as unicast (forward_routed()), ECMP(), or flood (DNC) over
       "uninteresting" sub-parts of the topology, e.g., parts where, for
       path steering reasons, different trees do not need to take
       different paths.

   The total number of bits to describe the topology vs. the number of
   BFERs in a BIFT:SI can range widely based on the size of the topology
   and the amount of alternative paths in it.  In a BIER-TE topology
   crafted by a BIER-TE expert, the higher the percentage of non-BFER
   bits, the higher the likelihood that those topology bits are not just
   BIER-TE overhead without additional benefit but instead will allow
   the expression of desirable path steering alternatives.

5.3.3.  Assigning BFR-ids with BIER-TE



   BIER-TE forwarding does not use BFR-ids, nor does it require that the
   BFIR-id field of the BIER header be set to a particular value.
   However, other parts of a BIER-TE deployment may need a BFR-id --
   specifically, multicast flow overlay signaling and multicast flow
   overlay packet disposition; in that case, BFRs need to also have BFR-
   ids for BIER-TE SDs.

   For example, for BIER overlay signaling, BFIRs need to have a BFR-id,
   because this BFIR BFR-id is carried in the BFIR-id field of the BIER
   header to indicate to the overlay signaling on the receiving BFER
   which BFIR originated the packet.

   In BIER, BFR-id = SI * BSL + BP, such that the SI and BP of a BFER
   can be calculated from the BFR-id and vice versa.  This also means
   that every BFR with a BFR-id has a reserved BP in an SI, even if that
   is not necessary for BIER forwarding, because the BFR may never be a
   BFER (i.e., will only be a BFIR).

   In BIER-TE, for a non-leaf BFER, there is usually a single BP for
   that BFER with a local_decap() adjacency on the BFER.  The BFR-id for
   such a BFER can therefore be determined using the same procedure as
   that used for (non-TE) BIER: BFR-id = SI * BSL + BP.

   As explained in Section 5.1.3, leaf BFERs do not need such a unique
   local_decap() adjacency.  Likewise, BFIRs that are not also BFERs may
   not have a unique local_decap() adjacency either.  For all those
   BFIRs and (leaf) BFERs, the controller needs to determine unique BFR-
   ids that do not collide with the BFR-ids derived from the non-leaf
   BFER local_decap() BPs.

   While this document defines no requirements on how to allocate such
   BFR-ids, a simple option is to derive it from the (SI,BP) of an
   adjacency that is unique to the BFR in question.  For a BFIR, this
   can be the first adjacency that is only populated on this BFIR; for a
   leaf BFER, this could be the first BP with an adjacency towards that
   BFER.

5.3.4.  Mapping from BFRs to BitStrings with BIER-TE



   In BIER, applications of the flow overlay on a BFIR can calculate the
   (SI,BP) of a BFER from the BFR-id of the BFER and can therefore
   easily determine the BitStrings for a BIER packet to a set of BFERs
   with known BFR-ids.

   In BIER-TE, this mapping needs to be equally supported for flow
   overlays.  This section outlines two core options, based on what type
   of Tree Engineering the BIER-TE controller needs to perform for a
   particular application.

   "Independent branches":  For a given flow overlay instance, the
      branches from a BFIR to every BFER are calculated by the BIER-TE
      controller to be independent of the branches to any other BFER.
      Shortest path trees are the most common examples of trees with
      independent branches.

   "Interdependent branches":  When a BFER is added to or deleted from a
      particular distribution tree, the BIER-TE controller has to
      recalculate the branches to other BFERs, because they may need to
      change.  Steiner trees are examples of interdependent branch
      trees.

   If "independent branches" are used, the BIER-TE controller can signal
   to the BFIR flow overlay for every BFER an SI:BitString that
   represents the branch to that BFER.  The flow overlay on the BFIR can
   then, independently of the controller, calculate the SI:BitString for
   all desired BFERs by ORing their BitStrings.  This allows flow
   overlay applications to operate independently of the controller
   whenever they need to determine which subset of BFERs needs to
   receive a particular packet.

   If "interdependent branches" are required, an application would need
   to query the SI:BitString for a given set of BFERs whenever the set
   changes.

   Note that in either case (unlike the scenario for BIER), the bits may
   need to change upon link/node failure/recovery, network expansion, or
   network resource consumption by other traffic as part of achieving
   Traffic Engineering goals (e.g., reoptimization of lower-priority
   traffic flows).  Interactions between such BFIR applications and the
   BIER-TE controller do therefore need to support dynamic updates to
   the SIs:BitStrings.

   Communications between the BFIR flow overlay and the BIER-TE
   controller require some way to identify the BFERs.  If BFR-ids are
   used in the deployment, as outlined in Section 5.3.3, then those are
   the "natural" BFR-ids.  If BFR-ids are not used, then any other
   unique identifier, such as a BFR's BFR-prefix [RFC8279], could be
   used.

5.3.5.  Assigning BFR-ids for BIER-TE



   It is not currently determined if a single subdomain could or should
   be allowed to forward both (non-TE) BIER and BIER-TE packets.  If
   this should be supported, there are two options:

   A.  BIER and BIER-TE have different BFR-ids in the same subdomain.
       This allows higher replication efficiency for BIER because the
       BIER BFR-ids can be assigned sequentially, while the BitStrings
       for BIER-TE will also have to assign the additional bits for the
       topology adjacencies.  There is no relationship between a BFR
       BIER BFR-id and its BIER-TE BFR-id.

   B.  BIER and BIER-TE share the same BFR-id.  The BFR-ids are assigned
       as explained above for BIER-TE and simply reused for BIER.  The
       replication efficiency for BIER will be as low as that for BIER-
       TE in this approach.

5.3.6.  Example Bit Allocations



5.3.6.1.  With BIER



   Consider a network setup with a BSL of 256 for a network topology as
   shown in Figure 17.  The network has six areas, each with 170 BFERs,
   connecting via a core with four (core) BFRs.  To address all BFERs
   with BIER, four SIs are required.  To send a BIER packet to all BFERs
   in the network, four copies need to be sent by the BFIR.  On the
   BFIR, it does not matter how the BFR-ids are allocated to BFERs in
   the network, but it does matter for efficiency further down in the
   network.

                area1           area2        area3
               BFR1a BFR1b  BFR2a BFR2b   BFR3a BFR3b
                 |  \         /    \        /  |
                 ................................
                 .                Core          .
                 ................................
                 |    /       \    /        \  |
               BFR4a BFR4b  BFR5a BFR5b   BFR6a BFR6b
                area4          area5        area6

                  Figure 17: Scaling BIER-TE Bits by Reuse

   With random allocation of BFR-ids to BFERs, each receiving area would
   (most likely) have to receive all four copies of the BIER packet
   because there would be BFR-ids for each of the four SIs in each of
   the areas.  Only further towards each BFER would this duplication
   subside -- when each of the four trees runs out of branches.

   If BFR-ids are allocated intelligently, then all the BFERs in an area
   would be given BFR-ids with as few different SIs as possible.  Each
   area would only have to forward one or two packets instead of four.

   Given how networks can grow over time, replication efficiency in an
   area will then also go down over time when BFR-ids are only allocated
   sequentially, network wide.  An area that initially only has BFR-ids
   in one SI might end up with many SIs over a longer period of growth.
   Allocating SIs to areas that initially have sufficiently many spare
   bits for growth can help alleviate this issue.  Alternatively, BFERs
   can be renumbered after network expansion.  In this example, one may
   consider using six SIs and assigning one to each area.

   This example shows that intelligent BFR-id allocation within at least
   subdomain 0 can be helpful or even necessary in BIER.

5.3.6.2.  With BIER-TE



   In BIER-TE, one needs to determine a subset of the physical topology
   and attached BFERs so that the "desired" representation of this
   topology and the BFERs fit into a single BitString.  This process
   needs to be repeated until the whole topology is covered.

   Once bits/SIs are assigned to the topology and BFERs, BFR-ids are
   just a derived set of identifiers from the operator / BIER-TE
   controller as explained above.

   Whenever different subtopologies have overlap, bits need to be
   repeated across the BitStrings, increasing the overall amount of bits
   required across all BitStrings/SIs.  In the worst case, one assigns
   random subsets of BFERs to different SIs.  This will result in an
   outcome much worse than in (non-TE) BIER: it maximizes the amount of
   unnecessary topology overlap across SIs and therefore reduces the
   number of BFERs that can be reached across each individual SI.
   Intelligent BFER-to-SI assignment and selecting specific "desired"
   subtopologies can minimize this problem.

   To set up BIER-TE efficiently for the topology shown in Figure 17,
   the following bit allocation method can be used.  This method can
   easily be expanded to other, similarly structured larger topologies.

   Each area is allocated one or more SIs, depending on the number of
   future expected BFERs and the number of bits required for the
   topology in the area.  In this example, six SIs are used, one per
   area.

   In addition, we use four bits in each SI:

   bia:  (b)it (i)ngress (a)

   bib:  (b)it (i)ngress (b)

   bea:  (b)it (e)gress (a)

   beb:  (b)it (e)gress (b)

   These bits will be used to pass BIER packets from any BFIR via any
   combination of ingress area a/b BFRs and egress area a/b BFRs into a
   specific target area.  These bits are then set up with the right
   forward_routed() adjacencies on the BFIRs and area edge BFRs as
   follows.

   On all BFIRs in an area, j|j=1...6, bia in each BIFT:SI is populated
   with the same forward_routed(BFRja) and bib with
   forward_routed(BFRjb).  On all area edge BFRs, bea in
   BIFT:SI=k|k=1...6 is populated with forward_routed(BFRka) and beb in
   BIFT:SI=k with forward_routed(BFRkb).

   For BIER-TE forwarding of a packet to a subset of BFERs across all
   areas, a BFIR would create at most six copies, with SI=1...SI=6.  In
   each packet, the BitString includes bits for one area and the BFERs
   in that area, plus the four bits to indicate whether to pass this
   packet via the ingress area a or b border BFR and the egress area a
   or b border BFR, therefore allowing path steering for those two
   "unicast" legs: 1) BFIR to ingress area edge and 2) core to egress
   area edge.  Replication only happens inside the egress areas.  For
   BFERs that are in the same area as the BFIR, these four bits are not
   used.

5.3.7.  Summary



   BIER-TE can, like BIER, support multiple SIs within a subdomain.
   This allows application of the mapping BFR-id = SI * BSL + BP.  This
   also permits the reuse of the BIER architecture concept of BFR-ids
   and, therefore, minimization of BIER-TE-specific functions in
   possible BIER layer control plane mechanisms with BIER-TE, including
   flow overlay methods and BIER header fields.

   The number of BFIRs/BFERs possible in a subdomain is smaller than in
   BIER because BIER-TE uses additional bits for the topology.

   Subdomains in BIER-TE can be used as they are in BIER to create more
   efficient replication to known subsets of BFERs.

   Assigning bits for BFERs intelligently into the right SI is more
   important in BIER-TE than in BIER because of replication efficiency
   and the overall amount of bits required.

6.  Security Considerations



   If "Encapsulation for Bit Index Explicit Replication (BIER) in MPLS
   and Non-MPLS Networks" [RFC8296] is used, its security considerations
   also apply to BIER-TE.

   The security considerations of "Multicast Using Bit Index Explicit
   Replication (BIER)" [RFC8279] also apply to BIER-TE, with the
   following overriding or additional considerations.

   BIER-TE forwarding explicitly supports unicast "tunneling" of BIER
   packets via forward_routed() adjacencies.  The BIER domain security
   model is based on a subset of interfaces on a BFR that connect to
   other BFRs of the same BIER domain.  For BIER-TE, this security model
   equally applies to such unicast "tunneled" BIER packets.  This not
   only includes the need to filter received unicast "tunneled" BIER
   packets to prohibit the injection of such "tunneled" BIER packets
   from outside the BIER domain but also the need to prohibit
   forward_routed() adjacencies from leaking BIER packets from the BIER
   domain.  It SHOULD be possible to configure interfaces to be part of
   a BIER domain solely for sending and receiving unicast "tunneled"
   BIER packets even if the interface cannot send/receive BIER
   encapsulated packets.

   In BIER, the standardized methods for the routing underlays are IGPs
   with extensions to distribute BFR-ids and BFR-prefixes.  [RFC8401]
   specifies the extensions for IS-IS, and [RFC8444] specifies the
   extensions for OSPF.  Attacking the protocols for the BIER routing
   underlay or (non-TE) BIER layer control plane, or the impairment of
   any BFRs in a domain, may lead to successful attacks against the
   information that BIER-TE learns from the routing protocol (routes,
   next hops, BFR-ids, ...), enabling DoS attacks against paths or the
   addressing (BFR-ids, BFR-prefixes) used by BIER.

   The reference model for the BIER-TE layer control plane is a BIER-TE
   controller.  When such a controller is used, the impairment of an
   individual BFR in a domain causes no impairment of the BIER-TE
   control plane on other BFRs.  If a routing protocol is used to
   support forward_routed() adjacencies, then this is still an attack
   vector as in BIER, but only for BIER-TE forward_routed() adjacencies
   and not other adjacencies.

   Whereas IGP routing protocols are most often not well secured through
   cryptographic authentication and confidentiality, communications
   between controllers and routers such as those to be considered for
   the BIER-TE controller / control plane can be, and are, much more
   commonly secured with those security properties -- for example, by
   using "Secure Shell" (SSH) [RFC4253] for NETCONF [RFC6242]; or via
   "Transport Layer Security" (TLS), such as [RFC8253] for PCEP
   [RFC5440] or [RFC7589] for NETCONF.  BIER-TE controllers SHOULD use
   security equal to or better than these mechanisms.

   When any of these security mechanisms/protocols are used for
   communications between a BIER-TE controller and BFRs, their security
   considerations apply to BIER-TE.  In addition, the security
   considerations of "A Path Computation Element (PCE)-Based
   Architecture" [RFC4655] apply.

   The most important attack vector in BIER-TE is misconfiguration,
   either on the BFRs themselves or via the BIER-TE controller.
   Forwarding entries with DNC could be set up to create persistent
   loops, in which packets only expire because of TTL.  To minimize the
   impact of such attacks (or, more likely, unintentional
   misconfiguration by operators and/or bad BIER-TE controller
   software), the BIER-TE forwarding rules are defined to be as strict
   in clearing bits as possible.  The clearing of all bits with an
   adjacency on a BFR prohibits a looping packet from creating
   additional packet amplification through the misconfigured loop on the
   packet's second time or subsequent times around the loop, because all
   relevant adjacency bits would have been cleared on the first round
   through the loop.  As a result, looping packets can occur in BIER-TE
   to the same degree as is possible with unintentional or malicious
   loops in the routing underlay with BIER, or even with unicast
   traffic.

   Deployments where BIER-TE would likely be beneficial may include
   operational models where actual configuration changes from the
   controller are only required during non-production phases of the
   network's life cycle, e.g., in embedded networks or in manufacturing
   networks during such activities as plant reworking or repairs.  In
   these types of deployments, configuration changes could be locked out
   when the network is in production state and could only be
   (re-)enabled through reverting the network/installation to non-
   production state.  Such security designs would not only allow a
   deployment to provide additional layers of protection against
   configuration attacks but would, first and foremost, protect the
   active production process from such configuration attacks.

7.  IANA Considerations



   This document has no IANA actions.

8.  References



8.1.  Normative References



   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119,
              DOI 10.17487/RFC2119, March 1997,
              <https://www.rfc-editor.org/info/rfc2119>.

   [RFC8174]  Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
              2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
              May 2017, <https://www.rfc-editor.org/info/rfc8174>.

   [RFC8279]  Wijnands, IJ., Ed., Rosen, E., Ed., Dolganow, A.,
              Przygienda, T., and S. Aldrin, "Multicast Using Bit Index
              Explicit Replication (BIER)", RFC 8279,
              DOI 10.17487/RFC8279, November 2017,
              <https://www.rfc-editor.org/info/rfc8279>.

   [RFC8296]  Wijnands, IJ., Ed., Rosen, E., Ed., Dolganow, A.,
              Tantsura, J., Aldrin, S., and I. Meilik, "Encapsulation
              for Bit Index Explicit Replication (BIER) in MPLS and Non-
              MPLS Networks", RFC 8296, DOI 10.17487/RFC8296, January
              2018, <https://www.rfc-editor.org/info/rfc8296>.

8.2.  Informative References



   [BIER-MCAST-OVERLAY]
              Trossen, D., Rahman, A., Wang, C., and T. Eckert,
              "Applicability of BIER Multicast Overlay for Adaptive
              Streaming Services", Work in Progress, Internet-Draft,
              draft-ietf-bier-multicast-http-response-06, 10 July 2021,
              <https://datatracker.ietf.org/doc/html/draft-ietf-bier-
              multicast-http-response-06>.

   [BIER-TE-PROTECTION]
              Eckert, T., Cauchie, G., Braun, W., and M. Menth,
              "Protection Methods for BIER-TE", Work in Progress,
              Internet-Draft, draft-eckert-bier-te-frr-03, 5 March 2018,
              <https://datatracker.ietf.org/doc/html/draft-eckert-bier-
              te-frr-03>.

   [BIER-TE-YANG]
              Zhang, Z., Wang, C., Chen, R., Hu, F., Sivakumar, M., and
              H. Chen, "A YANG data model for Tree Engineering for Bit
              Index Explicit Replication (BIER-TE)", Work in Progress,
              Internet-Draft, draft-ietf-bier-te-yang-05, 1 May 2022,
              <https://datatracker.ietf.org/doc/html/draft-ietf-bier-te-
              yang-05>.

   [Bloom70]  Bloom, B. H., "Space/time trade-offs in hash coding with
              allowable errors", Comm. ACM 13(7):422-6,
              DOI 10.1145/362686.362692, July 1970,
              <https://dl.acm.org/doi/10.1145/362686.362692>.

   [CONSTRAINED-CAST]
              Bergmann, O., Bormann, C., Gerdes, S., and H. Chen,
              "Constrained-Cast: Source-Routed Multicast for RPL", Work
              in Progress, Internet-Draft, draft-ietf-roll-ccast-01, 30
              October 2017, <https://datatracker.ietf.org/doc/html/
              draft-ietf-roll-ccast-01>.

   [ICC]      Reed, M. J., Al-Naday, M., Thomos, N., Trossen, D.,
              Petropoulos, G., and S. Spirou, "Stateless multicast
              switching in software defined networks", IEEE
              International Conference on Communications (ICC), Kuala
              Lumpur, Malaysia, DOI 10.1109/ICC.2016.7511036, May 2016,
              <https://ieeexplore.ieee.org/document/7511036>.

   [NON-MPLS-BIER-ENCODING]
              Wijnands, IJ., Mishra, M., Xu, X., and H. Bidgoli, "An
              Optional Encoding of the BIFT-id Field in the non-MPLS
              BIER Encapsulation", Work in Progress, Internet-Draft,
              draft-ietf-bier-non-mpls-bift-encoding-04, 30 May 2021,
              <https://datatracker.ietf.org/doc/html/draft-ietf-bier-
              non-mpls-bift-encoding-04>.

   [RCSD94]   Zhang, H. and D. Ferrari, "Rate-Controlled Service
              Disciplines", Journal of High Speed Networks, Volume 3,
              Issue 4, pp. 389-412, DOI 10.3233/JHS-1994-3405, October
              1994, <https://content.iospress.com/articles/journal-of-
              high-speed-networks/jhs3-4-05>.

   [RFC4253]  Ylonen, T. and C. Lonvick, Ed., "The Secure Shell (SSH)
              Transport Layer Protocol", RFC 4253, DOI 10.17487/RFC4253,
              January 2006, <https://www.rfc-editor.org/info/rfc4253>.

   [RFC4456]  Bates, T., Chen, E., and R. Chandra, "BGP Route
              Reflection: An Alternative to Full Mesh Internal BGP
              (IBGP)", RFC 4456, DOI 10.17487/RFC4456, April 2006,
              <https://www.rfc-editor.org/info/rfc4456>.

   [RFC4655]  Farrel, A., Vasseur, J.-P., and J. Ash, "A Path
              Computation Element (PCE)-Based Architecture", RFC 4655,
              DOI 10.17487/RFC4655, August 2006,
              <https://www.rfc-editor.org/info/rfc4655>.

   [RFC5440]  Vasseur, JP., Ed. and JL. Le Roux, Ed., "Path Computation
              Element (PCE) Communication Protocol (PCEP)", RFC 5440,
              DOI 10.17487/RFC5440, March 2009,
              <https://www.rfc-editor.org/info/rfc5440>.

   [RFC6241]  Enns, R., Ed., Bjorklund, M., Ed., Schoenwaelder, J., Ed.,
              and A. Bierman, Ed., "Network Configuration Protocol
              (NETCONF)", RFC 6241, DOI 10.17487/RFC6241, June 2011,
              <https://www.rfc-editor.org/info/rfc6241>.

   [RFC6242]  Wasserman, M., "Using the NETCONF Protocol over Secure
              Shell (SSH)", RFC 6242, DOI 10.17487/RFC6242, June 2011,
              <https://www.rfc-editor.org/info/rfc6242>.

   [RFC7589]  Badra, M., Luchuk, A., and J. Schoenwaelder, "Using the
              NETCONF Protocol over Transport Layer Security (TLS) with
              Mutual X.509 Authentication", RFC 7589,
              DOI 10.17487/RFC7589, June 2015,
              <https://www.rfc-editor.org/info/rfc7589>.

   [RFC7752]  Gredler, H., Ed., Medved, J., Previdi, S., Farrel, A., and
              S. Ray, "North-Bound Distribution of Link-State and
              Traffic Engineering (TE) Information Using BGP", RFC 7752,
              DOI 10.17487/RFC7752, March 2016,
              <https://www.rfc-editor.org/info/rfc7752>.

   [RFC7950]  Bjorklund, M., Ed., "The YANG 1.1 Data Modeling Language",
              RFC 7950, DOI 10.17487/RFC7950, August 2016,
              <https://www.rfc-editor.org/info/rfc7950>.

   [RFC7988]  Rosen, E., Ed., Subramanian, K., and Z. Zhang, "Ingress
              Replication Tunnels in Multicast VPN", RFC 7988,
              DOI 10.17487/RFC7988, October 2016,
              <https://www.rfc-editor.org/info/rfc7988>.

   [RFC8040]  Bierman, A., Bjorklund, M., and K. Watsen, "RESTCONF
              Protocol", RFC 8040, DOI 10.17487/RFC8040, January 2017,
              <https://www.rfc-editor.org/info/rfc8040>.

   [RFC8253]  Lopez, D., Gonzalez de Dios, O., Wu, Q., and D. Dhody,
              "PCEPS: Usage of TLS to Provide a Secure Transport for the
              Path Computation Element Communication Protocol (PCEP)",
              RFC 8253, DOI 10.17487/RFC8253, October 2017,
              <https://www.rfc-editor.org/info/rfc8253>.

   [RFC8345]  Clemm, A., Medved, J., Varga, R., Bahadur, N.,
              Ananthakrishnan, H., and X. Liu, "A YANG Data Model for
              Network Topologies", RFC 8345, DOI 10.17487/RFC8345, March
              2018, <https://www.rfc-editor.org/info/rfc8345>.

   [RFC8401]  Ginsberg, L., Ed., Przygienda, T., Aldrin, S., and Z.
              Zhang, "Bit Index Explicit Replication (BIER) Support via
              IS-IS", RFC 8401, DOI 10.17487/RFC8401, June 2018,
              <https://www.rfc-editor.org/info/rfc8401>.

   [RFC8402]  Filsfils, C., Ed., Previdi, S., Ed., Ginsberg, L.,
              Decraene, B., Litkowski, S., and R. Shakir, "Segment
              Routing Architecture", RFC 8402, DOI 10.17487/RFC8402,
              July 2018, <https://www.rfc-editor.org/info/rfc8402>.

   [RFC8444]  Psenak, P., Ed., Kumar, N., Wijnands, IJ., Dolganow, A.,
              Przygienda, T., Zhang, J., and S. Aldrin, "OSPFv2
              Extensions for Bit Index Explicit Replication (BIER)",
              RFC 8444, DOI 10.17487/RFC8444, November 2018,
              <https://www.rfc-editor.org/info/rfc8444>.

   [RFC8556]  Rosen, E., Ed., Sivakumar, M., Przygienda, T., Aldrin, S.,
              and A. Dolganow, "Multicast VPN Using Bit Index Explicit
              Replication (BIER)", RFC 8556, DOI 10.17487/RFC8556, April
              2019, <https://www.rfc-editor.org/info/rfc8556>.

   [TE-OVERVIEW]
              Farrel, A., Ed., "Overview and Principles of Internet
              Traffic Engineering", Work in Progress, Internet-Draft,
              draft-ietf-teas-rfc3272bis-21, 11 September 2022,
              <https://datatracker.ietf.org/doc/html/draft-ietf-teas-
              rfc3272bis-21>.

Appendix A.  BIER-TE and Segment Routing (SR)



   SR [RFC8402] aims to enable lightweight path steering via loose
   source routing.  For example, compared to its more heavyweight
   predecessor, RSVP-TE, SR does not require per-path signaling to each
   of these hops.

   BIER-TE supports the same design philosophy for multicast.  Like SR,
   BIER-TE

   *  relies on source routing (via a BitString), and

   *  only requires consideration of the "hops" either (1) on which
      replication has to happen or (2) across which the traffic should
      be steered (even without replication).

   Any other hops can be skipped via the use of routed adjacencies.

   BIER-TE "bit positions" (BPs) can be understood as the BIER-TE
   equivalent of "forwarding segments" in SR, but they have a different
   scope than do forwarding segments in SR.  Whereas forwarding segments
   in SR are global or local, BPs in BIER-TE have a scope that is
   comprised of one or more BFRs that have adjacencies for the BPs in
   their BIFTs.  These segments can be called "adjacency-scoped"
   forwarding segments.

   Adjacency scope could be global, but then every BFR would need an
   adjacency for a given BP -- for example, a forward_routed() adjacency
   with encapsulation to the global SR "Segment Identifier" (SID) of the
   destination.  Such a BP would always result in ingress replication,
   though (as in [RFC7988]).  The first BFR encountering this BP would
   directly replicate traffic on it.  Only by using non-global adjacency
   scope for BPs can traffic be steered and replicated on a non-BFIR.

   SR can naturally be combined with BIER-TE and can help optimize it.
   For example, instead of defining bit positions for non-replicating
   hops, it is equally possible to use SR encapsulations (e.g., SR-MPLS
   label stacks) for the encapsulation of "forward_routed()"
   adjacencies.

   Note that (non-TE) BIER itself can also be seen as being similar to
   SR.  BIER BPs act as global destination Node-SIDs, and the BIER
   BitString is simply a highly optimized mechanism to indicate multiple
   such SIDs and let the network take care of effectively replicating
   the packet hop by hop to each destination Node-SID.  BIER does not
   allow the indication of intermediate hops or, in terms of SR, the
   ability to indicate a sequence of SIDs to reach the destination.  On
   the other hand, BIER-TE and its adjacency-scoped BPs provide these
   capabilities.

Acknowledgements



   The authors would like to thank Greg Shepherd, IJsbrand Wijnands,
   Neale Ranns, Dirk Trossen, Sandy Zheng, Lou Berger, Jeffrey Zhang,
   Carsten Bormann, and Wolfgang Braun for their reviews and
   suggestions.

   Special thanks to Xuesong Geng for shepherding this document.
   Special thanks also for IESG review/suggestions by Alvaro Retana
   (responsible AD/RTG), Benjamin Kaduk (SEC), Tommy Pauly (TSV),
   Zaheduzzaman Sarker (TSV), Éric Vyncke (INT), Martin Vigoureux (RTG),
   Robert Wilton (OPS), Erik Kline (INT), Lars Eggert (GEN), Roman
   Danyliw (SEC), Ines Robles (RTGDIR), Robert Sparks (Gen-ART),
   Yingzhen Qu (RTGDIR), and Martin Duke (TSV).

Authors' Addresses



   Toerless Eckert (editor)
   Futurewei Technologies Inc.
   2330 Central Expy
   Santa Clara, CA 95050
   United States of America
   Email: tte@cs.fau.de


   Michael Menth
   University of Tuebingen
   Germany
   Email: menth@uni-tuebingen.de


   Gregory Cauchie
   KOEVOO
   France
   Email: gregory@koevoo.tech