RFC 2328




Network Working Group                                             J. Moy
Request for Comments: 2328                   Ascend Communications, Inc.
STD: 54                                                       April 1998
Obsoletes: 2178
Category: Standards Track


                             OSPF Version 2


Status of this Memo



    This document specifies an Internet standards track protocol for the
    Internet community, and requests discussion and suggestions for
    improvements.  Please refer to the current edition of the "Internet
    Official Protocol Standards" (STD 1) for the standardization state
    and status of this protocol.  Distribution of this memo is
    unlimited.

Copyright Notice



    Copyright (C) The Internet Society (1998).  All Rights Reserved.

Abstract



    This memo documents version 2 of the OSPF protocol.  OSPF is a
    link-state routing protocol.  It is designed to be run internal to a
    single Autonomous System.  Each OSPF router maintains an identical
    database describing the Autonomous System's topology.  From this
    database, a routing table is calculated by constructing a shortest-
    path tree.

    OSPF recalculates routes quickly in the face of topological changes,
    utilizing a minimum of routing protocol traffic.  OSPF provides
    support for equal-cost multipath.  An area routing capability is
    provided, enabling an additional level of routing protection and a
    reduction in routing protocol traffic.  In addition, all OSPF
    routing protocol exchanges are authenticated.

    The differences between this memo and RFC 2178 are explained in
    Appendix G. All differences are backward-compatible in nature.




Moy                         Standards Track                     [Page 1]

RFC 2328                     OSPF Version 2                   April 1998


    Implementations of this memo and of RFCs 2178, 1583, and 1247 will
    interoperate.

    Please send comments to ospf@gated.cornell.edu.

Table of Contents



    1        Introduction ........................................... 6
    1.1      Protocol Overview ...................................... 6
    1.2      Definitions of commonly used terms ..................... 8
    1.3      Brief history of link-state routing technology ........ 11
    1.4      Organization of this document ......................... 12
    1.5      Acknowledgments ....................................... 12
    2        The link-state database: organization and calculations  13
    2.1      Representation of routers and networks ................ 13
    2.1.1    Representation of non-broadcast networks .............. 15
    2.1.2    An example link-state database ........................ 18
    2.2      The shortest-path tree ................................ 21
    2.3      Use of external routing information ................... 23
    2.4      Equal-cost multipath .................................. 26
    3        Splitting the AS into Areas ........................... 26
    3.1      The backbone of the Autonomous System ................. 27
    3.2      Inter-area routing .................................... 27
    3.3      Classification of routers ............................. 28
    3.4      A sample area configuration ........................... 29
    3.5      IP subnetting support ................................. 35
    3.6      Supporting stub areas ................................. 37
    3.7      Partitions of areas ................................... 38
    4        Functional Summary .................................... 40
    4.1      Inter-area routing .................................... 41
    4.2      AS external routes .................................... 41
    4.3      Routing protocol packets .............................. 42
    4.4      Basic implementation requirements ..................... 43
    4.5      Optional OSPF capabilities ............................ 46
    5        Protocol data structures .............................. 47
    6        The Area Data Structure ............................... 49
    7        Bringing Up Adjacencies ............................... 52
    7.1      The Hello Protocol .................................... 52
    7.2      The Synchronization of Databases ...................... 53
    7.3      The Designated Router ................................. 54
    7.4      The Backup Designated Router .......................... 56
    7.5      The graph of adjacencies .............................. 56



Moy                         Standards Track                     [Page 2]

RFC 2328                     OSPF Version 2                   April 1998


    8        Protocol Packet Processing ............................ 58
    8.1      Sending protocol packets .............................. 58
    8.2      Receiving protocol packets ............................ 61
    9        The Interface Data Structure .......................... 63
    9.1      Interface states ...................................... 67
    9.2      Events causing interface state changes ................ 70
    9.3      The Interface state machine ........................... 72
    9.4      Electing the Designated Router ........................ 75
    9.5      Sending Hello packets ................................. 77
    9.5.1    Sending Hello packets on NBMA networks ................ 79
    10       The Neighbor Data Structure ........................... 80
    10.1     Neighbor states ....................................... 83
    10.2     Events causing neighbor state changes ................. 87
    10.3     The Neighbor state machine ............................ 89
    10.4     Whether to become adjacent ............................ 95
    10.5     Receiving Hello Packets ............................... 96
    10.6     Receiving Database Description Packets ................ 99
    10.7     Receiving Link State Request Packets ................. 102
    10.8     Sending Database Description Packets ................. 103
    10.9     Sending Link State Request Packets ................... 104
    10.10    An Example ........................................... 105
    11       The Routing Table Structure .......................... 107
    11.1     Routing table lookup ................................. 111
    11.2     Sample routing table, without areas .................. 111
    11.3     Sample routing table, with areas ..................... 112
    12       Link State Advertisements (LSAs) ..................... 115
    12.1     The LSA Header ....................................... 116
    12.1.1   LS age ............................................... 116
    12.1.2   Options .............................................. 117
    12.1.3   LS type .............................................. 117
    12.1.4   Link State ID ........................................ 117
    12.1.5   Advertising Router ................................... 119
    12.1.6   LS sequence number ................................... 120
    12.1.7   LS checksum .......................................... 121
    12.2     The link state database .............................. 121
    12.3     Representation of TOS ................................ 122
    12.4     Originating LSAs ..................................... 123
    12.4.1   Router-LSAs .......................................... 126
    12.4.1.1 Describing point-to-point interfaces ................. 130
    12.4.1.2 Describing broadcast and NBMA interfaces ............. 130
    12.4.1.3 Describing virtual links ............................. 131
    12.4.1.4 Describing Point-to-MultiPoint interfaces ............ 131



Moy                         Standards Track                     [Page 3]

RFC 2328                     OSPF Version 2                   April 1998


    12.4.1.5 Examples of router-LSAs .............................. 132
    12.4.2   Network-LSAs ......................................... 133
    12.4.2.1 Examples of network-LSAs ............................. 134
    12.4.3   Summary-LSAs ......................................... 135
    12.4.3.1 Originating summary-LSAs into stub areas ............. 137
    12.4.3.2 Examples of summary-LSAs ............................. 138
    12.4.4   AS-external-LSAs ..................................... 139
    12.4.4.1 Examples of AS-external-LSAs ......................... 140
    13       The Flooding Procedure ............................... 143
    13.1     Determining which LSA is newer ....................... 146
    13.2     Installing LSAs in the database ...................... 147
    13.3     Next step in the flooding procedure .................. 148
    13.4     Receiving self-originated LSAs ....................... 151
    13.5     Sending Link State Acknowledgment packets ............ 152
    13.6     Retransmitting LSAs .................................. 154
    13.7     Receiving link state acknowledgments ................. 155
    14       Aging The Link State Database ........................ 156
    14.1     Premature aging of LSAs .............................. 157
    15       Virtual Links ........................................ 158
    16       Calculation of the routing table ..................... 160
    16.1     Calculating the shortest-path tree for an area ....... 161
    16.1.1   The next hop calculation ............................. 167
    16.2     Calculating the inter-area routes .................... 178
    16.3     Examining transit areas' summary-LSAs ................ 170
    16.4     Calculating AS external routes ....................... 173
    16.4.1   External path preferences ............................ 175
    16.5     Incremental updates -- summary-LSAs .................. 175
    16.6     Incremental updates -- AS-external-LSAs .............. 177
    16.7     Events generated as a result of routing table changes  177
    16.8     Equal-cost multipath ................................. 178
             Footnotes ............................................ 179
             References ........................................... 183
    A        OSPF data formats .................................... 185
    A.1      Encapsulation of OSPF packets ........................ 185
    A.2      The Options field .................................... 187
    A.3      OSPF Packet Formats .................................. 189
    A.3.1    The OSPF packet header ............................... 190
    A.3.2    The Hello packet ..................................... 193
    A.3.3    The Database Description packet ...................... 195
    A.3.4    The Link State Request packet ........................ 197
    A.3.5    The Link State Update packet ......................... 199
    A.3.6    The Link State Acknowledgment packet ................. 201



Moy                         Standards Track                     [Page 4]

RFC 2328                     OSPF Version 2                   April 1998


    A.4      LSA formats .......................................... 203
    A.4.1    The LSA header ....................................... 204
    A.4.2    Router-LSAs .......................................... 206
    A.4.3    Network-LSAs ......................................... 210
    A.4.4    Summary-LSAs ......................................... 212
    A.4.5    AS-external-LSAs ..................................... 214
    B        Architectural Constants .............................. 217
    C        Configurable Constants ............................... 219
    C.1      Global parameters .................................... 219
    C.2      Area parameters ...................................... 220
    C.3      Router interface parameters .......................... 221
    C.4      Virtual link parameters .............................. 224
    C.5      NBMA network parameters .............................. 224
    C.6      Point-to-MultiPoint network parameters ............... 225
    C.7      Host route parameters ................................ 226
    D        Authentication ....................................... 227
    D.1      Null authentication .................................. 227
    D.2      Simple password authentication ....................... 228
    D.3      Cryptographic authentication ......................... 228
    D.4      Message generation ................................... 231
    D.4.1    Generating Null authentication ....................... 231
    D.4.2    Generating Simple password authentication ............ 232
    D.4.3    Generating Cryptographic authentication .............. 232
    D.5      Message verification ................................. 234
    D.5.1    Verifying Null authentication ........................ 234
    D.5.2    Verifying Simple password authentication ............. 234
    D.5.3    Verifying Cryptographic authentication ............... 235
    E        An algorithm for assigning Link State IDs ............ 236
    F        Multiple interfaces to the same network/subnet ....... 239
    G        Differences from RFC 2178 ............................ 240
    G.1      Flooding modifications ............................... 240
    G.2      Changes to external path preferences ................. 241
    G.3      Incomplete resolution of virtual next hops ........... 241
    G.4      Routing table lookup ................................. 241
             Security Considerations .............................. 243
             Author's Address ..................................... 243
             Full Copyright Statement ............................. 244








Moy                         Standards Track                     [Page 5]

RFC 2328                     OSPF Version 2                   April 1998


1.  Introduction



    This document is a specification of the Open Shortest Path First
    (OSPF) TCP/IP internet routing protocol.  OSPF is classified as an
    Interior Gateway Protocol (IGP).  This means that it distributes
    routing information between routers belonging to a single Autonomous
    System.  The OSPF protocol is based on link-state or SPF technology.
    This is a departure from the Bellman-Ford base used by traditional
    TCP/IP internet routing protocols.

    The OSPF protocol was developed by the OSPF working group of the
    Internet Engineering Task Force.  It has been designed expressly for
    the TCP/IP internet environment, including explicit support for CIDR
    and the tagging of externally-derived routing information.  OSPF
    also provides for the authentication of routing updates, and
    utilizes IP multicast when sending/receiving the updates.  In
    addition, much work has been done to produce a protocol that
    responds quickly to topology changes, yet involves small amounts of
    routing protocol traffic.

    1.1.  Protocol overview



        OSPF routes IP packets based solely on the destination IP
        address found in the IP packet header.  IP packets are routed
        "as is" -- they are not encapsulated in any further protocol
        headers as they transit the Autonomous System.  OSPF is a
        dynamic routing protocol.  It quickly detects topological
        changes in the AS (such as router interface failures) and
        calculates new loop-free routes after a period of convergence.
        This period of convergence is short and involves a minimum of
        routing traffic.

        In a link-state routing protocol, each router maintains a
        database describing the Autonomous System's topology.  This
        database is referred to as the link-state database. Each
        participating router has an identical database.  Each individual
        piece of this database is a particular router's local state
        (e.g., the router's usable interfaces and reachable neighbors).
        The router distributes its local state throughout the Autonomous
        System by flooding.





Moy                         Standards Track                     [Page 6]

RFC 2328                     OSPF Version 2                   April 1998


        All routers run the exact same algorithm, in parallel.  From the
        link-state database, each router constructs a tree of shortest
        paths with itself as root.  This shortest-path tree gives the
        route to each destination in the Autonomous System.  Externally
        derived routing information appears on the tree as leaves.

        When several equal-cost routes to a destination exist, traffic
        is distributed equally among them.  The cost of a route is
        described by a single dimensionless metric.

        OSPF allows sets of networks to be grouped together.  Such a
        grouping is called an area.  The topology of an area is hidden
        from the rest of the Autonomous System.  This information hiding
        enables a significant reduction in routing traffic.  Also,
        routing within the area is determined only by the area's own
        topology, lending the area protection from bad routing data.  An
        area is a generalization of an IP subnetted network.

        OSPF enables the flexible configuration of IP subnets.  Each
        route distributed by OSPF has a destination and mask.  Two
        different subnets of the same IP network number may have
        different sizes (i.e., different masks).  This is commonly
        referred to as variable length subnetting.  A packet is routed
        to the best (i.e., longest or most specific) match.  Host routes
        are considered to be subnets whose masks are "all ones"
        (0xffffffff).

        All OSPF protocol exchanges are authenticated.  This means that
        only trusted routers can participate in the Autonomous System's
        routing.  A variety of authentication schemes can be used; in
        fact, separate authentication schemes can be configured for each
        IP subnet.

        Externally derived routing data (e.g., routes learned from an
        Exterior Gateway Protocol such as BGP; see [Ref23]) is
        advertised throughout the Autonomous System.  This externally
        derived data is kept separate from the OSPF protocol's link
        state data.  Each external route can also be tagged by the
        advertising router, enabling the passing of additional
        information between routers on the boundary of the Autonomous
        System.




Moy                         Standards Track                     [Page 7]

RFC 2328                     OSPF Version 2                   April 1998


    1.2.  Definitions of commonly used terms



        This section provides definitions for terms that have a specific
        meaning to the OSPF protocol and that are used throughout the
        text.  The reader unfamiliar with the Internet Protocol Suite is
        referred to [Ref13] for an introduction to IP.


        Router
            A level three Internet Protocol packet switch.  Formerly
            called a gateway in much of the IP literature.

        Autonomous System
            A group of routers exchanging routing information via a
            common routing protocol.  Abbreviated as AS.

        Interior Gateway Protocol
            The routing protocol spoken by the routers belonging to an
            Autonomous system.  Abbreviated as IGP.  Each Autonomous
            System has a single IGP.  Separate Autonomous Systems may be
            running different IGPs.

        Router ID
            A 32-bit number assigned to each router running the OSPF
            protocol.  This number uniquely identifies the router within
            an Autonomous System.

        Network
            In this memo, an IP network/subnet/supernet.  It is possible
            for one physical network to be assigned multiple IP
            network/subnet numbers.  We consider these to be separate
            networks.  Point-to-point physical networks are an exception
            - they are considered a single network no matter how many
            (if any at all) IP network/subnet numbers are assigned to
            them.

        Network mask
            A 32-bit number indicating the range of IP addresses
            residing on a single IP network/subnet/supernet.  This
            specification displays network masks as hexadecimal numbers.





Moy                         Standards Track                     [Page 8]

RFC 2328                     OSPF Version 2                   April 1998


            For example, the network mask for a class C IP network is
            displayed as 0xffffff00.  Such a mask is often displayed
            elsewhere in the literature as 255.255.255.0.

        Point-to-point networks
            A network that joins a single pair of routers.  A 56Kb
            serial line is an example of a point-to-point network.

        Broadcast networks
            Networks supporting many (more than two) attached routers,
            together with the capability to address a single physical
            message to all of the attached routers (broadcast).
            Neighboring routers are discovered dynamically on these nets
            using OSPF's Hello Protocol.  The Hello Protocol itself
            takes advantage of the broadcast capability.  The OSPF
            protocol makes further use of multicast capabilities, if
            they exist.  Each pair of routers on a broadcast network is
            assumed to be able to communicate directly. An ethernet is
            an example of a broadcast network.

        Non-broadcast networks
            Networks supporting many (more than two) routers, but having
            no broadcast capability.  Neighboring routers are maintained
            on these nets using OSPF's Hello Protocol.  However, due to
            the lack of broadcast capability, some configuration
            information may be necessary to aid in the discovery of
            neighbors.  On non-broadcast networks, OSPF protocol packets
            that are normally multicast need to be sent to each
            neighboring router, in turn. An X.25 Public Data Network
            (PDN) is an example of a non-broadcast network.

            OSPF runs in one of two modes over non-broadcast networks.
            The first mode, called non-broadcast multi-access or NBMA,
            simulates the operation of OSPF on a broadcast network. The
            second mode, called Point-to-MultiPoint, treats the non-
            broadcast network as a collection of point-to-point links.
            Non-broadcast networks are referred to as NBMA networks or
            Point-to-MultiPoint networks, depending on OSPF's mode of
            operation over the network.






Moy                         Standards Track                     [Page 9]

RFC 2328                     OSPF Version 2                   April 1998


        Interface
            The connection between a router and one of its attached
            networks.  An interface has state information associated
            with it, which is obtained from the underlying lower level
            protocols and the routing protocol itself.  An interface to
            a network has associated with it a single IP address and
            mask (unless the network is an unnumbered point-to-point
            network).  An interface is sometimes also referred to as a
            link.

        Neighboring routers
            Two routers that have interfaces to a common network.
            Neighbor relationships are maintained by, and usually
            dynamically discovered by, OSPF's Hello Protocol.

        Adjacency
            A relationship formed between selected neighboring routers
            for the purpose of exchanging routing information.  Not
            every pair of neighboring routers become adjacent.

        Link state advertisement
            Unit of data describing the local state of a router or
            network. For a router, this includes the state of the
            router's interfaces and adjacencies.  Each link state
            advertisement is flooded throughout the routing domain. The
            collected link state advertisements of all routers and
            networks forms the protocol's link state database.
            Throughout this memo, link state advertisement is
            abbreviated as LSA.

        Hello Protocol
            The part of the OSPF protocol used to establish and maintain
            neighbor relationships.  On broadcast networks the Hello
            Protocol can also dynamically discover neighboring routers.

        Flooding
            The part of the OSPF protocol that distributes and
            synchronizes the link-state database between OSPF routers.

        Designated Router
            Each broadcast and NBMA network that has at least two
            attached routers has a Designated Router.  The Designated



Moy                         Standards Track                    [Page 10]

RFC 2328                     OSPF Version 2                   April 1998


            Router generates an LSA for the network and has other
            special responsibilities in the running of the protocol.
            The Designated Router is elected by the Hello Protocol.

            The Designated Router concept enables a reduction in the
            number of adjacencies required on a broadcast or NBMA
            network.  This in turn reduces the amount of routing
            protocol traffic and the size of the link-state database.

        Lower-level protocols
            The underlying network access protocols that provide
            services to the Internet Protocol and in turn the OSPF
            protocol.  Examples of these are the X.25 packet and frame
            levels for X.25 PDNs, and the ethernet data link layer for
            ethernets.


    1.3.  Brief history of link-state routing technology



        OSPF is a link state routing protocol.  Such protocols are also
        referred to in the literature as SPF-based or distributed-
        database protocols.  This section gives a brief description of
        the developments in link-state technology that have influenced
        the OSPF protocol.

        The first link-state routing protocol was developed for use in
        the ARPANET packet switching network.  This protocol is
        described in [Ref3].  It has formed the starting point for all
        other link-state protocols.  The homogeneous ARPANET
        environment, i.e., single-vendor packet switches connected by
        synchronous serial lines, simplified the design and
        implementation of the original protocol.

        Modifications to this protocol were proposed in [Ref4].  These
        modifications dealt with increasing the fault tolerance of the
        routing protocol through, among other things, adding a checksum
        to the LSAs (thereby detecting database corruption).  The paper
        also included means for reducing the routing traffic overhead in
        a link-state protocol.  This was accomplished by introducing
        mechanisms which enabled the interval between LSA originations
        to be increased by an order of magnitude.




Moy                         Standards Track                    [Page 11]

RFC 2328                     OSPF Version 2                   April 1998


        A link-state algorithm has also been proposed for use as an ISO
        IS-IS routing protocol.  This protocol is described in [Ref2].
        The protocol includes methods for data and routing traffic
        reduction when operating over broadcast networks.  This is
        accomplished by election of a Designated Router for each
        broadcast network, which then originates an LSA for the network.

        The OSPF Working Group of the IETF has extended this work in
        developing the OSPF protocol.  The Designated Router concept has
        been greatly enhanced to further reduce the amount of routing
        traffic required.  Multicast capabilities are utilized for
        additional routing bandwidth reduction.  An area routing scheme
        has been developed enabling information
        hiding/protection/reduction.  Finally, the algorithms have been
        tailored for efficient operation in TCP/IP internets.


    1.4.  Organization of this document



        The first three sections of this specification give a general
        overview of the protocol's capabilities and functions.  Sections
        4-16 explain the protocol's mechanisms in detail.  Packet
        formats, protocol constants and configuration items are
        specified in the appendices.

        Labels such as HelloInterval encountered in the text refer to
        protocol constants.  They may or may not be configurable.
        Architectural constants are summarized in Appendix B.
        Configurable constants are summarized in Appendix C.

        The detailed specification of the protocol is presented in terms
        of data structures.  This is done in order to make the
        explanation more precise.  Implementations of the protocol are
        required to support the functionality described, but need not
        use the precise data structures that appear in this memo.


    1.5.  Acknowledgments



        The author would like to thank Ran Atkinson, Fred Baker, Jeffrey
        Burgan, Rob Coltun, Dino Farinacci, Vince Fuller, Phanindra
        Jujjavarapu, Milo Medin, Tom Pusateri, Kannan Varadhan, Zhaohui



Moy                         Standards Track                    [Page 12]

RFC 2328                     OSPF Version 2                   April 1998


        Zhang and the rest of the OSPF Working Group for the ideas and
        support they have given to this project.

        The OSPF Point-to-MultiPoint interface is based on work done by
        Fred Baker.

        The OSPF Cryptographic Authentication option was developed by
        Fred Baker and Ran Atkinson.


2.  The Link-state Database: organization and calculations



    The following subsections describe the organization of OSPF's link-
    state database, and the routing calculations that are performed on
    the database in order to produce a router's routing table.


    2.1.  Representation of routers and networks



        The Autonomous System's link-state database describes a directed
        graph.  The vertices of the graph consist of routers and
        networks.  A graph edge connects two routers when they are
        attached via a physical point-to-point network.  An edge
        connecting a router to a network indicates that the router has
        an interface on the network. Networks can be either transit or
        stub networks. Transit networks are those capable of carrying
        data traffic that is neither locally originated nor locally
        destined. A transit network is represented by a graph vertex
        having both incoming and outgoing edges. A stub network's vertex
        has only incoming edges.

        The neighborhood of each network node in the graph depends on
        the network's type (point-to-point, broadcast, NBMA or Point-
        to-MultiPoint) and the number of routers having an interface to
        the network.  Three cases are depicted in Figure 1a.  Rectangles
        indicate routers.  Circles and oblongs indicate networks.
        Router names are prefixed with the letters RT and network names
        with the letter N.  Router interface names are prefixed by the
        letter I.  Lines between routers indicate point-to-point
        networks.  The left side of the figure shows networks with their
        connected routers, with the resulting graphs shown on the right.




Moy                         Standards Track                    [Page 13]

RFC 2328                     OSPF Version 2                   April 1998





                                                  **FROM**

                                           *      |RT1|RT2|
                +---+Ia    +---+           *   ------------
                |RT1|------|RT2|           T   RT1|   | X |
                +---+    Ib+---+           O   RT2| X |   |
                                           *    Ia|   | X |
                                           *    Ib| X |   |

                     Physical point-to-point networks


                                                  **FROM**
                      +---+                *
                      |RT7|                *      |RT7| N3|
                      +---+                T   ------------
                        |                  O   RT7|   |   |
            +----------------------+       *    N3| X |   |
                       N3                  *

                              Stub networks

                                                  **FROM**
                +---+      +---+
                |RT3|      |RT4|              |RT3|RT4|RT5|RT6|N2 |
                +---+      +---+        *  ------------------------
                  |    N2    |          *  RT3|   |   |   |   | X |
            +----------------------+    T  RT4|   |   |   |   | X |
                  |          |          O  RT5|   |   |   |   | X |
                +---+      +---+        *  RT6|   |   |   |   | X |
                |RT5|      |RT6|        *   N2| X | X | X | X |   |
                +---+      +---+

                          Broadcast or NBMA networks



                    Figure 1a: Network map components




Moy                         Standards Track                    [Page 14]

RFC 2328                     OSPF Version 2                   April 1998


             Networks and routers are represented by vertices.
             An edge connects Vertex A to Vertex B iff the
             intersection of Column A and Row B is marked with
                                  an X.



        The top of Figure 1a shows two routers connected by a point-to-
        point link. In the resulting link-state database graph, the two
        router vertices are directly connected by a pair of edges, one
        in each direction. Interfaces to point-to-point networks need
        not be assigned IP addresses.  When interface addresses are
        assigned, they are modelled as stub links, with each router
        advertising a stub connection to the other router's interface
        address. Optionally, an IP subnet can be assigned to the point-
        to-point network. In this case, both routers advertise a stub
        link to the IP subnet, instead of advertising each others' IP
        interface addresses.

        The middle of Figure 1a shows a network with only one attached
        router (i.e., a stub network). In this case, the network appears
        on the end of a stub connection in the link-state database's
        graph.

        When multiple routers are attached to a broadcast network, the
        link-state database graph shows all routers bidirectionally
        connected to the network vertex. This is pictured at the bottom
        of Figure 1a.

        Each network (stub or transit) in the graph has an IP address
        and associated network mask.  The mask indicates the number of
        nodes on the network.  Hosts attached directly to routers
        (referred to as host routes) appear on the graph as stub
        networks.  The network mask for a host route is always
        0xffffffff, which indicates the presence of a single node.


        2.1.1.  Representation of non-broadcast networks



            As mentioned previously, OSPF can run over non-broadcast
            networks in one of two modes: NBMA or Point-to-MultiPoint.
            The choice of mode determines the way that the Hello



Moy                         Standards Track                    [Page 15]

RFC 2328                     OSPF Version 2                   April 1998


            protocol and flooding work over the non-broadcast network,
            and the way that the network is represented in the link-
            state database.

            In NBMA mode, OSPF emulates operation over a broadcast
            network: a Designated Router is elected for the NBMA
            network, and the Designated Router originates an LSA for the
            network. The graph representation for broadcast networks and
            NBMA networks is identical. This representation is pictured
            in the middle of Figure 1a.

            NBMA mode is the most efficient way to run OSPF over non-
            broadcast networks, both in terms of link-state database
            size and in terms of the amount of routing protocol traffic.
            However, it has one significant restriction: it requires all
            routers attached to the NBMA network to be able to
            communicate directly. This restriction may be met on some
            non-broadcast networks, such as an ATM subnet utilizing
            SVCs. But it is often not met on other non-broadcast
            networks, such as PVC-only Frame Relay networks. On non-
            broadcast networks where not all routers can communicate
            directly you can break the non-broadcast network into
            logical subnets, with the routers on each subnet being able
            to communicate directly, and then run each separate subnet
            as an NBMA network (see [Ref15]). This however requires
            quite a bit of administrative overhead, and is prone to
            misconfiguration. It is probably better to run such a non-
            broadcast network in Point-to-Multipoint mode.

            In Point-to-MultiPoint mode, OSPF treats all router-to-
            router connections over the non-broadcast network as if they
            were point-to-point links. No Designated Router is elected
            for the network, nor is there an LSA generated for the
            network. In fact, a vertex for the Point-to-MultiPoint
            network does not appear in the graph of the link-state
            database.

            Figure 1b illustrates the link-state database representation
            of a Point-to-MultiPoint network. On the left side of the
            figure, a Point-to-MultiPoint network is pictured. It is
            assumed that all routers can communicate directly, except
            for routers RT4 and RT5. I3 though I6 indicate the routers'



Moy                         Standards Track                    [Page 16]

RFC 2328                     OSPF Version 2                   April 1998


            IP interface addresses on the Point-to-MultiPoint network.
            In the graphical representation of the link-state database,
            routers that can communicate directly over the Point-to-
            MultiPoint network are joined by bidirectional edges, and
            each router also has a stub connection to its own IP
            interface address (which is in contrast to the
            representation of real point-to-point links; see Figure 1a).

            On some non-broadcast networks, use of Point-to-MultiPoint
            mode and data-link protocols such as Inverse ARP (see
            [Ref14]) will allow autodiscovery of OSPF neighbors even
            though broadcast support is not available.






                                                  **FROM**
                +---+      +---+
                |RT3|      |RT4|              |RT3|RT4|RT5|RT6|
                +---+      +---+        *  --------------------
                I3|    N2    |I4        *  RT3|   | X | X | X |
            +----------------------+    T  RT4| X |   |   | X |
                I5|          |I6        O  RT5| X |   |   | X |
                +---+      +---+        *  RT6| X | X | X |   |
                |RT5|      |RT6|        *   I3| X |   |   |   |
                +---+      +---+            I4|   | X |   |   |
                                            I5|   |   | X |   |
                                            I6|   |   |   | X |



                    Figure 1b: Network map components
                       Point-to-MultiPoint networks

             All routers can communicate directly over N2, except
                routers RT4 and RT5. I3 through I6 indicate IP
                           interface addresses






Moy                         Standards Track                    [Page 17]

RFC 2328                     OSPF Version 2                   April 1998


        2.1.2.  An example link-state database



            Figure 2 shows a sample map of an Autonomous System.  The
            rectangle labelled H1 indicates a host, which has a SLIP
            connection to Router RT12.  Router RT12 is therefore
            advertising a host route.  Lines between routers indicate
            physical point-to-point networks.  The only point-to-point
            network that has been assigned interface addresses is the
            one joining Routers RT6 and RT10.  Routers RT5 and RT7 have
            BGP connections to other Autonomous Systems.  A set of BGP-
            learned routes have been displayed for both of these
            routers.

            A cost is associated with the output side of each router
            interface.  This cost is configurable by the system
            administrator.  The lower the cost, the more likely the
            interface is to be used to forward data traffic.  Costs are
            also associated with the externally derived routing data
            (e.g., the BGP-learned routes).

            The directed graph resulting from the map in Figure 2 is
            depicted in Figure 3.  Arcs are labelled with the cost of
            the corresponding router output interface.  Arcs having no
            labelled cost have a cost of 0.  Note that arcs leading from
            networks to routers always have cost 0; they are significant
            nonetheless.  Note also that the externally derived routing
            data appears on the graph as stubs.

            The link-state database is pieced together from LSAs
            generated by the routers.  In the associated graphical
            representation, the neighborhood of each router or transit
            network is represented in a single, separate LSA.  Figure 4
            shows these LSAs graphically. Router RT12 has an interface
            to two broadcast networks and a SLIP line to a host.
            Network N6 is a broadcast network with three attached
            routers.  The cost of all links from Network N6 to its
            attached routers is 0.  Note that the LSA for Network N6 is
            actually generated by one of the network's attached routers:
            the router that has been elected Designated Router for the
            network.





Moy                         Standards Track                    [Page 18]

RFC 2328                     OSPF Version 2                   April 1998



                 +
                 | 3+---+                     N12      N14
               N1|--|RT1|\ 1                    \ N13 /
                 |  +---+ \                     8\ |8/8
                 +         \ ____                 \|/
                            /    \   1+---+8    8+---+6
                           *  N3  *---|RT4|------|RT5|--------+
                            \____/    +---+      +---+        |
                  +         /   |                  |7         |
                  | 3+---+ /    |                  |          |
                N2|--|RT2|/1    |1                 |6         |
                  |  +---+    +---+8            6+---+        |
                  +           |RT3|--------------|RT6|        |
                              +---+              +---+        |
                                |2               Ia|7         |
                                |                  |          |
                           +---------+             |          |
                               N4                  |          |
                                                   |          |
                                                   |          |
                       N11                         |          |
                   +---------+                     |          |
                        |                          |          |    N12
                        |3                         |          |6 2/
                      +---+                        |        +---+/
                      |RT9|                        |        |RT7|---N15
                      +---+                        |        +---+ 9
                        |1                   +     |          |1
                       _|__                  |   Ib|5       __|_
                      /    \      1+----+2   |  3+----+1   /    \
                     *  N9  *------|RT11|----|---|RT10|---*  N6  *
                      \____/       +----+    |   +----+    \____/
                        |                    |                |
                        |1                   +                |1
             +--+   10+----+                N8              +---+
             |H1|-----|RT12|                                |RT8|
             +--+SLIP +----+                                +---+
                        |2                                    |4
                        |                                     |
                   +---------+                            +--------+
                       N10                                    N7



Moy                         Standards Track                    [Page 19]

RFC 2328                     OSPF Version 2                   April 1998


                    Figure 2: A sample Autonomous System

                                **FROM**

                 |RT|RT|RT|RT|RT|RT|RT|RT|RT|RT|RT|RT|
                 |1 |2 |3 |4 |5 |6 |7 |8 |9 |10|11|12|N3|N6|N8|N9|
              ----- ---------------------------------------------
              RT1|  |  |  |  |  |  |  |  |  |  |  |  |0 |  |  |  |
              RT2|  |  |  |  |  |  |  |  |  |  |  |  |0 |  |  |  |
              RT3|  |  |  |  |  |6 |  |  |  |  |  |  |0 |  |  |  |
              RT4|  |  |  |  |8 |  |  |  |  |  |  |  |0 |  |  |  |
              RT5|  |  |  |8 |  |6 |6 |  |  |  |  |  |  |  |  |  |
              RT6|  |  |8 |  |7 |  |  |  |  |5 |  |  |  |  |  |  |
              RT7|  |  |  |  |6 |  |  |  |  |  |  |  |  |0 |  |  |
          *   RT8|  |  |  |  |  |  |  |  |  |  |  |  |  |0 |  |  |
          *   RT9|  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |0 |
          T  RT10|  |  |  |  |  |7 |  |  |  |  |  |  |  |0 |0 |  |
          O  RT11|  |  |  |  |  |  |  |  |  |  |  |  |  |  |0 |0 |
          *  RT12|  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |0 |
          *    N1|3 |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |
               N2|  |3 |  |  |  |  |  |  |  |  |  |  |  |  |  |  |
               N3|1 |1 |1 |1 |  |  |  |  |  |  |  |  |  |  |  |  |
               N4|  |  |2 |  |  |  |  |  |  |  |  |  |  |  |  |  |
               N6|  |  |  |  |  |  |1 |1 |  |1 |  |  |  |  |  |  |
               N7|  |  |  |  |  |  |  |4 |  |  |  |  |  |  |  |  |
               N8|  |  |  |  |  |  |  |  |  |3 |2 |  |  |  |  |  |
               N9|  |  |  |  |  |  |  |  |1 |  |1 |1 |  |  |  |  |
              N10|  |  |  |  |  |  |  |  |  |  |  |2 |  |  |  |  |
              N11|  |  |  |  |  |  |  |  |3 |  |  |  |  |  |  |  |
              N12|  |  |  |  |8 |  |2 |  |  |  |  |  |  |  |  |  |
              N13|  |  |  |  |8 |  |  |  |  |  |  |  |  |  |  |  |
              N14|  |  |  |  |8 |  |  |  |  |  |  |  |  |  |  |  |
              N15|  |  |  |  |  |  |9 |  |  |  |  |  |  |  |  |  |
               H1|  |  |  |  |  |  |  |  |  |  |  |10|  |  |  |  |


                     Figure 3: The resulting directed graph

                 Networks and routers are represented by vertices.
                 An edge of cost X connects Vertex A to Vertex B iff
                 the intersection of Column A and Row B is marked
                                     with an X.



Moy                         Standards Track                    [Page 20]

RFC 2328                     OSPF Version 2                   April 1998


                     **FROM**                       **FROM**

                  |RT12|N9|N10|H1|                 |RT9|RT11|RT12|N9|
           *  --------------------          *  ----------------------
           *  RT12|    |  |   |  |          *   RT9|   |    |    |0 |
           T    N9|1   |  |   |  |          T  RT11|   |    |    |0 |
           O   N10|2   |  |   |  |          O  RT12|   |    |    |0 |
           *    H1|10  |  |   |  |          *    N9|   |    |    |  |
           *                                *
                RT12's router-LSA              N9's network-LSA

                  Figure 4: Individual link state components

              Networks and routers are represented by vertices.
              An edge of cost X connects Vertex A to Vertex B iff
              the intersection of Column A and Row B is marked
                                  with an X.

    2.2.  The shortest-path tree



        When no OSPF areas are configured, each router in the Autonomous
        System has an identical link-state database, leading to an
        identical graphical representation.  A router generates its
        routing table from this graph by calculating a tree of shortest
        paths with the router itself as root.  Obviously, the shortest-
        path tree depends on the router doing the calculation.  The
        shortest-path tree for Router RT6 in our example is depicted in
        Figure 5.

        The tree gives the entire path to any destination network or
        host.  However, only the next hop to the destination is used in
        the forwarding process.  Note also that the best route to any
        router has also been calculated.  For the processing of external
        data, we note the next hop and distance to any router
        advertising external routes.  The resulting routing table for
        Router RT6 is pictured in Table 2.  Note that there is a
        separate route for each end of a numbered point-to-point network
        (in this case, the serial line between Routers RT6 and RT10).


        Routes to networks belonging to other AS'es (such as N12) appear
        as dashed lines on the shortest path tree in Figure 5.  Use of



Moy                         Standards Track                    [Page 21]

RFC 2328                     OSPF Version 2                   April 1998



                                RT6(origin)
                    RT5 o------------o-----------o Ib
                       /|\    6      |\     7
                     8/8|8\          | \
                     /  |  \        6|  \
                    o   |   o        |   \7
                   N12  o  N14       |    \
                       N13        2  |     \
                            N4 o-----o RT3  \
                                    /        \    5
                                  1/     RT10 o-------o Ia
                                  /           |\
                       RT4 o-----o N3        3| \1
                                /|            |  \ N6     RT7
                               / |         N8 o   o---------o
                              /  |            |   |        /|
                         RT2 o   o RT1        |   |      2/ |9
                            /    |            |   |RT8   /  |
                           /3    |3      RT11 o   o     o   o
                          /      |            |   |    N12 N15
                      N2 o       o N1        1|   |4
                                              |   |
                                           N9 o   o N7
                                             /|
                                            / |
                        N11      RT9       /  |RT12
                         o--------o-------o   o--------o H1
                             3                |   10
                                              |2
                                              |
                                              o N10


                     Figure 5: The SPF tree for Router RT6

              Edges that are not marked with a cost have a cost of
              of zero (these are network-to-router links). Routes
              to networks N12-N15 are external information that is
                         considered in Section 2.3





Moy                         Standards Track                    [Page 22]

RFC 2328                     OSPF Version 2                   April 1998


                   Destination   Next  Hop   Distance
                   __________________________________
                   N1            RT3         10
                   N2            RT3         10
                   N3            RT3         7
                   N4            RT3         8
                   Ib            *           7
                   Ia            RT10        12
                   N6            RT10        8
                   N7            RT10        12
                   N8            RT10        10
                   N9            RT10        11
                   N10           RT10        13
                   N11           RT10        14
                   H1            RT10        21
                   __________________________________
                   RT5           RT5         6
                   RT7           RT10        8


    Table 2: The portion of Router RT6's routing table listing local
                             destinations.

        this externally derived routing information is considered in the
        next section.


    2.3.  Use of external routing information



        After the tree is created the external routing information is
        examined.  This external routing information may originate from
        another routing protocol such as BGP, or be statically
        configured (static routes).  Default routes can also be included
        as part of the Autonomous System's external routing information.

        External routing information is flooded unaltered throughout the
        AS.  In our example, all the routers in the Autonomous System
        know that Router RT7 has two external routes, with metrics 2 and
        9.

        OSPF supports two types of external metrics.  Type 1 external
        metrics are expressed in the same units as OSPF interface cost



Moy                         Standards Track                    [Page 23]

RFC 2328                     OSPF Version 2                   April 1998


        (i.e., in terms of the link state metric).  Type 2 external
        metrics are an order of magnitude larger; any Type 2 metric is
        considered greater than the cost of any path internal to the AS.
        Use of Type 2 external metrics assumes that routing between
        AS'es is the major cost of routing a packet, and eliminates the
        need for conversion of external costs to internal link state
        metrics.

        As an example of Type 1 external metric processing, suppose that
        the Routers RT7 and RT5 in Figure 2 are advertising Type 1
        external metrics.  For each advertised external route, the total
        cost from Router RT6 is calculated as the sum of the external
        route's advertised cost and the distance from Router RT6 to the
        advertising router.  When two routers are advertising the same
        external destination, RT6 picks the advertising router providing
        the minimum total cost. RT6 then sets the next hop to the
        external destination equal to the next hop that would be used
        when routing packets to the chosen advertising router.

        In Figure 2, both Router RT5 and RT7 are advertising an external
        route to destination Network N12.  Router RT7 is preferred since
        it is advertising N12 at a distance of 10 (8+2) to Router RT6,
        which is better than Router RT5's 14 (6+8).  Table 3 shows the
        entries that are added to the routing table when external routes
        are examined:



                         Destination   Next  Hop   Distance
                         __________________________________
                         N12           RT10        10
                         N13           RT5         14
                         N14           RT5         14
                         N15           RT10        17


                 Table 3: The portion of Router RT6's routing table
                           listing external destinations.


        Processing of Type 2 external metrics is simpler.  The AS
        boundary router advertising the smallest external metric is



Moy                         Standards Track                    [Page 24]

RFC 2328                     OSPF Version 2                   April 1998


        chosen, regardless of the internal distance to the AS boundary
        router.  Suppose in our example both Router RT5 and Router RT7
        were advertising Type 2 external routes.  Then all traffic
        destined for Network N12 would be forwarded to Router RT7, since
        2 < 8.  When several equal-cost Type 2 routes exist, the
        internal distance to the advertising routers is used to break
        the tie.

        Both Type 1 and Type 2 external metrics can be present in the AS
        at the same time.  In that event, Type 1 external metrics always
        take precedence.

        This section has assumed that packets destined for external
        destinations are always routed through the advertising AS
        boundary router.  This is not always desirable.  For example,
        suppose in Figure 2 there is an additional router attached to
        Network N6, called Router RTX.  Suppose further that RTX does
        not participate in OSPF routing, but does exchange BGP
        information with the AS boundary router RT7.  Then, Router RT7
        would end up advertising OSPF external routes for all
        destinations that should be routed to RTX.  An extra hop will
        sometimes be introduced if packets for these destinations need
        always be routed first to Router RT7 (the advertising router).

        To deal with this situation, the OSPF protocol allows an AS
        boundary router to specify a "forwarding address" in its AS-
        external-LSAs.  In the above example, Router RT7 would specify
        RTX's IP address as the "forwarding address" for all those
        destinations whose packets should be routed directly to RTX.

        The "forwarding address" has one other application.  It enables
        routers in the Autonomous System's interior to function as
        "route servers".  For example, in Figure 2 the router RT6 could
        become a route server, gaining external routing information
        through a combination of static configuration and external
        routing protocols.  RT6 would then start advertising itself as
        an AS boundary router, and would originate a collection of OSPF
        AS-external-LSAs.  In each AS-external-LSA, Router RT6 would
        specify the correct Autonomous System exit point to use for the
        destination through appropriate setting of the LSA's "forwarding
        address" field.




Moy                         Standards Track                    [Page 25]

RFC 2328                     OSPF Version 2                   April 1998


    2.4.  Equal-cost multipath



        The above discussion has been simplified by considering only a
        single route to any destination.  In reality, if multiple
        equal-cost routes to a destination exist, they are all
        discovered and used.  This requires no conceptual changes to the
        algorithm, and its discussion is postponed until we consider the
        tree-building process in more detail.

        With equal cost multipath, a router potentially has several
        available next hops towards any given destination.


3.  Splitting the AS into Areas



    OSPF allows collections of contiguous networks and hosts to be
    grouped together.  Such a group, together with the routers having
    interfaces to any one of the included networks, is called an area.
    Each area runs a separate copy of the basic link-state routing
    algorithm.  This means that each area has its own link-state
    database and corresponding graph, as explained in the previous
    section.

    The topology of an area is invisible from the outside of the area.
    Conversely, routers internal to a given area know nothing of the
    detailed topology external to the area.  This isolation of knowledge
    enables the protocol to effect a marked reduction in routing traffic
    as compared to treating the entire Autonomous System as a single
    link-state domain.

    With the introduction of areas, it is no longer true that all
    routers in the AS have an identical link-state database.  A router
    actually has a separate link-state database for each area it is
    connected to.  (Routers connected to multiple areas are called area
    border routers).  Two routers belonging to the same area have, for
    that area, identical area link-state databases.

    Routing in the Autonomous System takes place on two levels,
    depending on whether the source and destination of a packet reside
    in the same area (intra-area routing is used) or different areas
    (inter-area routing is used).  In intra-area routing, the packet is
    routed solely on information obtained within the area; no routing



Moy                         Standards Track                    [Page 26]

RFC 2328                     OSPF Version 2                   April 1998


    information obtained from outside the area can be used.  This
    protects intra-area routing from the injection of bad routing
    information.  We discuss inter-area routing in Section 3.2.


    3.1.  The backbone of the Autonomous System



        The OSPF backbone is the special OSPF Area 0 (often written as
        Area 0.0.0.0, since OSPF Area ID's are typically formatted as IP
        addresses). The OSPF backbone always contains all area border
        routers. The backbone is responsible for distributing routing
        information between non-backbone areas. The backbone must be
        contiguous. However, it need not be physically contiguous;
        backbone connectivity can be established/maintained through the
        configuration of virtual links.

        Virtual links can be configured between any two backbone routers
        that have an interface to a common non-backbone area.  Virtual
        links belong to the backbone.  The protocol treats two routers
        joined by a virtual link as if they were connected by an
        unnumbered point-to-point backbone network.  On the graph of the
        backbone, two such routers are joined by arcs whose costs are
        the intra-area distances between the two routers.  The routing
        protocol traffic that flows along the virtual link uses intra-
        area routing only.


    3.2.  Inter-area routing



        When routing a packet between two non-backbone areas the
        backbone is used.  The path that the packet will travel can be
        broken up into three contiguous pieces: an intra-area path from
        the source to an area border router, a backbone path between the
        source and destination areas, and then another intra-area path
        to the destination.  The algorithm finds the set of such paths
        that have the smallest cost.

        Looking at this another way, inter-area routing can be pictured
        as forcing a star configuration on the Autonomous System, with
        the backbone as hub and each of the non-backbone areas as
        spokes.




Moy                         Standards Track                    [Page 27]

RFC 2328                     OSPF Version 2                   April 1998


        The topology of the backbone dictates the backbone paths used
        between areas.  The topology of the backbone can be enhanced by
        adding virtual links.  This gives the system administrator some
        control over the routes taken by inter-area traffic.

        The correct area border router to use as the packet exits the
        source area is chosen in exactly the same way routers
        advertising external routes are chosen.  Each area border router
        in an area summarizes for the area its cost to all networks
        external to the area.  After the SPF tree is calculated for the
        area, routes to all inter-area destinations are calculated by
        examining the summaries of the area border routers.


    3.3.  Classification of routers



        Before the introduction of areas, the only OSPF routers having a
        specialized function were those advertising external routing
        information, such as Router RT5 in Figure 2.  When the AS is
        split into OSPF areas, the routers are further divided according
        to function into the following four overlapping categories:


        Internal routers
            A router with all directly connected networks belonging to
            the same area. These routers run a single copy of the basic
            routing algorithm.

        Area border routers
            A router that attaches to multiple areas.  Area border
            routers run multiple copies of the basic algorithm, one copy
            for each attached area. Area border routers condense the
            topological information of their attached areas for
            distribution to the backbone.  The backbone in turn
            distributes the information to the other areas.

        Backbone routers
            A router that has an interface to the backbone area.  This
            includes all routers that interface to more than one area
            (i.e., area border routers).  However, backbone routers do
            not have to be area border routers.  Routers with all
            interfaces connecting to the backbone area are supported.



Moy                         Standards Track                    [Page 28]

RFC 2328                     OSPF Version 2                   April 1998


        AS boundary routers
            A router that exchanges routing information with routers
            belonging to other Autonomous Systems.  Such a router
            advertises AS external routing information throughout the
            Autonomous System.  The paths to each AS boundary router are
            known by every router in the AS.  This classification is
            completely independent of the previous classifications: AS
            boundary routers may be internal or area border routers, and
            may or may not participate in the backbone.


    3.4.  A sample area configuration



        Figure 6 shows a sample area configuration.  The first area
        consists of networks N1-N4, along with their attached routers
        RT1-RT4.  The second area consists of networks N6-N8, along with
        their attached routers RT7, RT8, RT10 and RT11.  The third area
        consists of networks N9-N11 and Host H1, along with their
        attached routers RT9, RT11 and RT12.  The third area has been
        configured so that networks N9-N11 and Host H1 will all be
        grouped into a single route, when advertised external to the
        area (see Section 3.5 for more details).

        In Figure 6, Routers RT1, RT2, RT5, RT6, RT8, RT9 and RT12 are
        internal routers.  Routers RT3, RT4, RT7, RT10 and RT11 are area
        border routers.  Finally, as before, Routers RT5 and RT7 are AS
        boundary routers.

        Figure 7 shows the resulting link-state database for the Area 1.
        The figure completely describes that area's intra-area routing.
        It also shows the complete view of the internet for the two
        internal routers RT1 and RT2.  It is the job of the area border
        routers, RT3 and RT4, to advertise into Area 1 the distances to
        all destinations external to the area.  These are indicated in
        Figure 7 by the dashed stub routes.  Also, RT3 and RT4 must
        advertise into Area 1 the location of the AS boundary routers
        RT5 and RT7.  Finally, AS-external-LSAs from RT5 and RT7 are
        flooded throughout the entire AS, and in particular throughout
        Area 1.  These LSAs are included in Area 1's database, and yield
        routes to Networks N12-N15.

        Routers RT3 and RT4 must also summarize Area 1's topology for



Moy                         Standards Track                    [Page 29]

RFC 2328                     OSPF Version 2                   April 1998



             ...........................
             .   +                     .
             .   | 3+---+              .      N12      N14
             . N1|--|RT1|\ 1           .        \ N13 /
             .   |  +---+ \            .        8\ |8/8
             .   +         \ ____      .          \|/
             .              /    \   1+---+8    8+---+6
             .             *  N3  *---|RT4|------|RT5|--------+
             .              \____/    +---+      +---+        |
             .    +         /      \   .           |7         |
             .    | 3+---+ /        \  .           |          |
             .  N2|--|RT2|/1        1\ .           |6         |
             .    |  +---+            +---+8    6+---+        |
             .    +                   |RT3|------|RT6|        |
             .                        +---+      +---+        |
             .                      2/ .         Ia|7         |
             .                      /  .           |          |
             .             +---------+ .           |          |
             .Area 1           N4      .           |          |
             ...........................           |          |
          ..........................               |          |
          .            N11         .               |          |
          .        +---------+     .               |          |
          .             |          .               |          |    N12
          .             |3         .             Ib|5         |6 2/
          .           +---+        .             +----+     +---+/
          .           |RT9|        .    .........|RT10|.....|RT7|---N15.
          .           +---+        .    .        +----+     +---+ 9    .
          .             |1         .    .    +  /3    1\      |1       .
          .            _|__        .    .    | /        \   __|_       .
          .           /    \      1+----+2   |/          \ /    \      .
          .          *  N9  *------|RT11|----|            *  N6  *     .
          .           \____/       +----+    |             \____/      .
          .             |          .    .    |                |        .
          .             |1         .    .    +                |1       .
          .  +--+   10+----+       .    .   N8              +---+      .
          .  |H1|-----|RT12|       .    .                   |RT8|      .
          .  +--+SLIP +----+       .    .                   +---+      .
          .             |2         .    .                     |4       .
          .             |          .    .                     |        .
          .        +---------+     .    .                 +--------+   .



Moy                         Standards Track                    [Page 30]

RFC 2328                     OSPF Version 2                   April 1998


          .            N10         .    .                     N7       .
          .                        .    .Area 2                        .
          .Area 3                  .    ................................
          ..........................

                    Figure 6: A sample OSPF area configuration

        distribution to the backbone.  Their backbone LSAs are shown in
        Table 4.  These summaries show which networks are contained in
        Area 1 (i.e., Networks N1-N4), and the distance to these
        networks from the routers RT3 and RT4 respectively.


        The link-state database for the backbone is shown in Figure 8.
        The set of routers pictured are the backbone routers.  Router
        RT11 is a backbone router because it belongs to two areas.  In
        order to make the backbone connected, a virtual link has been
        configured between Routers R10 and R11.

        The area border routers RT3, RT4, RT7, RT10 and RT11 condense
        the routing information of their attached non-backbone areas for
        distribution via the backbone; these are the dashed stubs that
        appear in Figure 8.  Remember that the third area has been
        configured to condense Networks N9-N11 and Host H1 into a single
        route.  This yields a single dashed line for networks N9-N11 and
        Host H1 in Figure 8.  Routers RT5 and RT7 are AS boundary
        routers; their externally derived information also appears on
        the graph in Figure 8 as stubs.



                     Network   RT3 adv.   RT4 adv.
                     _____________________________
                     N1        4          4
                     N2        4          4
                     N3        1          1
                     N4        2          3

              Table 4: Networks advertised to the backbone
                        by Routers RT3 and RT4.





Moy                         Standards Track                    [Page 31]

RFC 2328                     OSPF Version 2                   April 1998



                               **FROM**

                          |RT|RT|RT|RT|RT|RT|
                          |1 |2 |3 |4 |5 |7 |N3|
                       ----- -------------------
                       RT1|  |  |  |  |  |  |0 |
                       RT2|  |  |  |  |  |  |0 |
                       RT3|  |  |  |  |  |  |0 |
                   *   RT4|  |  |  |  |  |  |0 |
                   *   RT5|  |  |14|8 |  |  |  |
                   T   RT7|  |  |20|14|  |  |  |
                   O    N1|3 |  |  |  |  |  |  |
                   *    N2|  |3 |  |  |  |  |  |
                   *    N3|1 |1 |1 |1 |  |  |  |
                        N4|  |  |2 |  |  |  |  |
                     Ia,Ib|  |  |20|27|  |  |  |
                        N6|  |  |16|15|  |  |  |
                        N7|  |  |20|19|  |  |  |
                        N8|  |  |18|18|  |  |  |
                 N9-N11,H1|  |  |29|36|  |  |  |
                       N12|  |  |  |  |8 |2 |  |
                       N13|  |  |  |  |8 |  |  |
                       N14|  |  |  |  |8 |  |  |
                       N15|  |  |  |  |  |9 |  |

                      Figure 7: Area 1's Database.

              Networks and routers are represented by vertices.
              An edge of cost X connects Vertex A to Vertex B iff
              the intersection of Column A and Row B is marked
                               with an X.













Moy                         Standards Track                    [Page 32]

RFC 2328                     OSPF Version 2                   April 1998


                                  **FROM**

                            |RT|RT|RT|RT|RT|RT|RT
                            |3 |4 |5 |6 |7 |10|11|
                         ------------------------
                         RT3|  |  |  |6 |  |  |  |
                         RT4|  |  |8 |  |  |  |  |
                         RT5|  |8 |  |6 |6 |  |  |
                         RT6|8 |  |7 |  |  |5 |  |
                         RT7|  |  |6 |  |  |  |  |
                     *  RT10|  |  |  |7 |  |  |2 |
                     *  RT11|  |  |  |  |  |3 |  |
                     T    N1|4 |4 |  |  |  |  |  |
                     O    N2|4 |4 |  |  |  |  |  |
                     *    N3|1 |1 |  |  |  |  |  |
                     *    N4|2 |3 |  |  |  |  |  |
                          Ia|  |  |  |  |  |5 |  |
                          Ib|  |  |  |7 |  |  |  |
                          N6|  |  |  |  |1 |1 |3 |
                          N7|  |  |  |  |5 |5 |7 |
                          N8|  |  |  |  |4 |3 |2 |
                   N9-N11,H1|  |  |  |  |  |  |11|
                         N12|  |  |8 |  |2 |  |  |
                         N13|  |  |8 |  |  |  |  |
                         N14|  |  |8 |  |  |  |  |
                         N15|  |  |  |  |9 |  |  |


                     Figure 8: The backbone's database.

              Networks and routers are represented by vertices.
              An edge of cost X connects Vertex A to Vertex B iff
              the intersection of Column A and Row B is marked
                                 with an X.

        The backbone enables the exchange of summary information between
        area border routers.  Every area border router hears the area
        summaries from all other area border routers.  It then forms a
        picture of the distance to all networks outside of its area by
        examining the collected LSAs, and adding in the backbone
        distance to each advertising router.




Moy                         Standards Track                    [Page 33]

RFC 2328                     OSPF Version 2                   April 1998


        Again using Routers RT3 and RT4 as an example, the procedure
        goes as follows: They first calculate the SPF tree for the
        backbone.  This gives the distances to all other area border
        routers.  Also noted are the distances to networks (Ia and Ib)
        and AS boundary routers (RT5 and RT7) that belong to the
        backbone.  This calculation is shown in Table 5.


        Next, by looking at the area summaries from these area border
        routers, RT3 and RT4 can determine the distance to all networks
        outside their area.  These distances are then advertised
        internally to the area by RT3 and RT4.  The advertisements that
        Router RT3 and RT4 will make into Area 1 are shown in Table 6.
        Note that Table 6 assumes that an area range has been configured
        for the backbone which groups Ia and Ib into a single LSA.


        The information imported into Area 1 by Routers RT3 and RT4
        enables an internal router, such as RT1, to choose an area
        border router intelligently.  Router RT1 would use RT4 for
        traffic to Network N6, RT3 for traffic to Network N10, and would


                              dist  from   dist  from
                              RT3          RT4
                   __________________________________
                   to  RT3    *            21
                   to  RT4    22           *
                   to  RT7    20           14
                   to  RT10   15           22
                   to  RT11   18           25
                   __________________________________
                   to  Ia     20           27
                   to  Ib     15           22
                   __________________________________
                   to  RT5    14           8
                   to  RT7    20           14

                 Table 5: Backbone distances calculated
                        by Routers RT3 and RT4.





Moy                         Standards Track                    [Page 34]

RFC 2328                     OSPF Version 2                   April 1998




                   Destination   RT3 adv.   RT4 adv.
                   _________________________________
                   Ia,Ib         20         27
                   N6            16         15
                   N7            20         19
                   N8            18         18
                   N9-N11,H1     29         36
                   _________________________________
                   RT5           14         8
                   RT7           20         14

              Table 6: Destinations advertised into Area 1
                        by Routers RT3 and RT4.

        load share between the two for traffic to Network N8.

        Router RT1 can also determine in this manner the shortest path
        to the AS boundary routers RT5 and RT7.  Then, by looking at RT5
        and RT7's AS-external-LSAs, Router RT1 can decide between RT5 or
        RT7 when sending to a destination in another Autonomous System
        (one of the networks N12-N15).

        Note that a failure of the line between Routers RT6 and RT10
        will cause the backbone to become disconnected.  Configuring a
        virtual link between Routers RT7 and RT10 will give the backbone
        more connectivity and more resistance to such failures.


    3.5.  IP subnetting support



        OSPF attaches an IP address mask to each advertised route.  The
        mask indicates the range of addresses being described by the
        particular route.  For example, a summary-LSA for the
        destination 128.185.0.0 with a mask of 0xffff0000 actually is
        describing a single route to the collection of destinations
        128.185.0.0 - 128.185.255.255.  Similarly, host routes are
        always advertised with a mask of 0xffffffff, indicating the
        presence of only a single destination.





Moy                         Standards Track                    [Page 35]

RFC 2328                     OSPF Version 2                   April 1998


        Including the mask with each advertised destination enables the
        implementation of what is commonly referred to as variable-
        length subnetting.  This means that a single IP class A, B, or C
        network number can be broken up into many subnets of various
        sizes.  For example, the network 128.185.0.0 could be broken up
        into 62 variable-sized subnets: 15 subnets of size 4K, 15
        subnets of size 256, and 32 subnets of size 8.  Table 7 shows
        some of the resulting network addresses together with their
        masks.



                  Network address   IP address mask   Subnet size
                  _______________________________________________
                  128.185.16.0      0xfffff000        4K
                  128.185.1.0       0xffffff00        256
                  128.185.0.8       0xfffffff8        8


                         Table 7: Some sample subnet sizes.


        There are many possible ways of dividing up a class A, B, and C
        network into variable sized subnets.  The precise procedure for
        doing so is beyond the scope of this specification.  This
        specification however establishes the following guideline: When
        an IP packet is forwarded, it is always forwarded to the network
        that is the best match for the packet's destination.  Here best
        match is synonymous with the longest or most specific match.
        For example, the default route with destination of 0.0.0.0 and
        mask 0x00000000 is always a match for every IP destination.  Yet
        it is always less specific than any other match.  Subnet masks
        must be assigned so that the best match for any IP destination
        is unambiguous.

        Attaching an address mask to each route also enables the support
        of IP supernetting. For example, a single physical network
        segment could be assigned the [address,mask] pair
        [192.9.4.0,0xfffffc00]. The segment would then be single IP
        network, containing addresses from the four consecutive class C
        network numbers 192.9.4.0 through 192.9.7.0. Such addressing is
        now becoming commonplace with the advent of CIDR (see [Ref10]).



Moy                         Standards Track                    [Page 36]

RFC 2328                     OSPF Version 2                   April 1998


        In order to get better aggregation at area boundaries, area
        address ranges can be employed (see Section C.2 for more
        details).  Each address range is defined as an [address,mask]
        pair.  Many separate networks may then be contained in a single
        address range, just as a subnetted network is composed of many
        separate subnets.  Area border routers then summarize the area
        contents (for distribution to the backbone) by advertising a
        single route for each address range.  The cost of the route is
        the maximum cost to any of the networks falling in the specified
        range.

        For example, an IP subnetted network might be configured as a
        single OSPF area.  In that case, a single address range could be
        configured:  a class A, B, or C network number along with its
        natural IP mask.  Inside the area, any number of variable sized
        subnets could be defined.  However, external to the area a
        single route for the entire subnetted network would be
        distributed, hiding even the fact that the network is subnetted
        at all.  The cost of this route is the maximum of the set of
        costs to the component subnets.


    3.6.  Supporting stub areas



        In some Autonomous Systems, the majority of the link-state
        database may consist of AS-external-LSAs.  An OSPF AS-external-
        LSA is usually flooded throughout the entire AS.  However, OSPF
        allows certain areas to be configured as "stub areas".  AS-
        external-LSAs are not flooded into/throughout stub areas;
        routing to AS external destinations in these areas is based on a
        (per-area) default only.  This reduces the link-state database
        size, and therefore the memory requirements, for a stub area's
        internal routers.

        In order to take advantage of the OSPF stub area support,
        default routing must be used in the stub area.  This is
        accomplished as follows.  One or more of the stub area's area
        border routers must advertise a default route into the stub area
        via summary-LSAs.  These summary defaults are flooded throughout
        the stub area, but no further.  (For this reason these defaults
        pertain only to the particular stub area).  These summary
        default routes will be used for any destination that is not



Moy                         Standards Track                    [Page 37]

RFC 2328                     OSPF Version 2                   April 1998


        explicitly reachable by an intra-area or inter-area path (i.e.,
        AS external destinations).

        An area can be configured as a stub when there is a single exit
        point from the area, or when the choice of exit point need not
        be made on a per-external-destination basis.  For example, Area
        3 in Figure 6 could be configured as a stub area, because all
        external traffic must travel though its single area border
        router RT11.  If Area 3 were configured as a stub, Router RT11
        would advertise a default route for distribution inside Area 3
        (in a summary-LSA), instead of flooding the AS-external-LSAs for
        Networks N12-N15 into/throughout the area.

        The OSPF protocol ensures that all routers belonging to an area
        agree on whether the area has been configured as a stub.  This
        guarantees that no confusion will arise in the flooding of AS-
        external-LSAs.

        There are a couple of restrictions on the use of stub areas.
        Virtual links cannot be configured through stub areas.  In
        addition, AS boundary routers cannot be placed internal to stub
        areas.


    3.7.  Partitions of areas



        OSPF does not actively attempt to repair area partitions.  When
        an area becomes partitioned, each component simply becomes a
        separate area.  The backbone then performs routing between the
        new areas.  Some destinations reachable via intra-area routing
        before the partition will now require inter-area routing.

        However, in order to maintain full routing after the partition,
        an address range must not be split across multiple components of
        the area partition. Also, the backbone itself must not
        partition.  If it does, parts of the Autonomous System will
        become unreachable.  Backbone partitions can be repaired by
        configuring virtual links (see Section 15).

        Another way to think about area partitions is to look at the
        Autonomous System graph that was introduced in Section 2.  Area
        IDs can be viewed as colors for the graph's edges.[1] Each edge



Moy                         Standards Track                    [Page 38]

RFC 2328                     OSPF Version 2                   April 1998


        of the graph connects to a network, or is itself a point-to-
        point network.  In either case, the edge is colored with the
        network's Area ID.

        A group of edges, all having the same color, and interconnected
        by vertices, represents an area.  If the topology of the
        Autonomous System is intact, the graph will have several regions
        of color, each color being a distinct Area ID.

        When the AS topology changes, one of the areas may become
        partitioned.  The graph of the AS will then have multiple
        regions of the same color (Area ID).  The routing in the
        Autonomous System will continue to function as long as these
        regions of same color are connected by the single backbone
        region.






























Moy                         Standards Track                    [Page 39]

RFC 2328                     OSPF Version 2                   April 1998


4.  Functional Summary



    A separate copy of OSPF's basic routing algorithm runs in each area.
    Routers having interfaces to multiple areas run multiple copies of
    the algorithm.  A brief summary of the routing algorithm follows.

    When a router starts, it first initializes the routing protocol data
    structures.  The router then waits for indications from the lower-
    level protocols that its interfaces are functional.

    A router then uses the OSPF's Hello Protocol to acquire neighbors.
    The router sends Hello packets to its neighbors, and in turn
    receives their Hello packets.  On broadcast and point-to-point
    networks, the router dynamically detects its neighboring routers by
    sending its Hello packets to the multicast address AllSPFRouters.
    On non-broadcast networks, some configuration information may be
    necessary in order to discover neighbors.  On broadcast and NBMA
    networks the Hello Protocol also elects a Designated router for the
    network.

    The router will attempt to form adjacencies with some of its newly
    acquired neighbors.  Link-state databases are synchronized between
    pairs of adjacent routers.  On broadcast and NBMA networks, the
    Designated Router determines which routers should become adjacent.

    Adjacencies control the distribution of routing information.
    Routing updates are sent and received only on adjacencies.

    A router periodically advertises its state, which is also called
    link state.  Link state is also advertised when a router's state
    changes.  A router's adjacencies are reflected in the contents of
    its LSAs.  This relationship between adjacencies and link state
    allows the protocol to detect dead routers in a timely fashion.

    LSAs are flooded throughout the area.  The flooding algorithm is
    reliable, ensuring that all routers in an area have exactly the same
    link-state database.  This database consists of the collection of
    LSAs originated by each router belonging to the area.  From this
    database each router calculates a shortest-path tree, with itself as
    root.  This shortest-path tree in turn yields a routing table for
    the protocol.




Moy                         Standards Track                    [Page 40]

RFC 2328                     OSPF Version 2                   April 1998


    4.1.  Inter-area routing



        The previous section described the operation of the protocol
        within a single area.  For intra-area routing, no other routing
        information is pertinent.  In order to be able to route to
        destinations outside of the area, the area border routers inject
        additional routing information into the area.  This additional
        information is a distillation of the rest of the Autonomous
        System's topology.

        This distillation is accomplished as follows: Each area border
        router is by definition connected to the backbone.  Each area
        border router summarizes the topology of its attached non-
        backbone areas for transmission on the backbone, and hence to
        all other area border routers.  An area border router then has
        complete topological information concerning the backbone, and
        the area summaries from each of the other area border routers.
        From this information, the router calculates paths to all
        inter-area destinations.  The router then advertises these paths
        into its attached areas.  This enables the area's internal
        routers to pick the best exit router when forwarding traffic
        inter-area destinations.


    4.2.  AS external routes



        Routers that have information regarding other Autonomous Systems
        can flood this information throughout the AS.  This external
        routing information is distributed verbatim to every
        participating router.  There is one exception: external routing
        information is not flooded into "stub" areas (see Section 3.6).

        To utilize external routing information, the path to all routers
        advertising external information must be known throughout the AS
        (excepting the stub areas).  For that reason, the locations of
        these AS boundary routers are summarized by the (non-stub) area
        border routers.








Moy                         Standards Track                    [Page 41]

RFC 2328                     OSPF Version 2                   April 1998


    4.3.  Routing protocol packets



        The OSPF protocol runs directly over IP, using IP protocol 89.
        OSPF does not provide any explicit fragmentation/reassembly
        support.  When fragmentation is necessary, IP
        fragmentation/reassembly is used.  OSPF protocol packets have
        been designed so that large protocol packets can generally be
        split into several smaller protocol packets.  This practice is
        recommended; IP fragmentation should be avoided whenever
        possible.

        Routing protocol packets should always be sent with the IP TOS
        field set to 0.  If at all possible, routing protocol packets
        should be given preference over regular IP data traffic, both
        when being sent and received.  As an aid to accomplishing this,
        OSPF protocol packets should have their IP precedence field set
        to the value Internetwork Control (see [Ref5]).

        All OSPF protocol packets share a common protocol header that is
        described in Appendix A.  The OSPF packet types are listed below
        in Table 8.  Their formats are also described in Appendix A.



             Type   Packet  name           Protocol  function
             __________________________________________________________
             1      Hello                  Discover/maintain  neighbors
             2      Database Description   Summarize database contents
             3      Link State Request     Database download
             4      Link State Update      Database update
             5      Link State Ack         Flooding acknowledgment


                            Table 8: OSPF packet types.


        OSPF's Hello protocol uses Hello packets to discover and
        maintain neighbor relationships.  The Database Description and
        Link State Request packets are used in the forming of
        adjacencies.  OSPF's reliable update mechanism is implemented by
        the Link State Update and Link State Acknowledgment packets.




Moy                         Standards Track                    [Page 42]

RFC 2328                     OSPF Version 2                   April 1998


        Each Link State Update packet carries a set of new link state
        advertisements (LSAs) one hop further away from their point of
        origination.  A single Link State Update packet may contain the
        LSAs of several routers.  Each LSA is tagged with the ID of the
        originating router and a checksum of its link state contents.
        Each LSA also has a type field; the different types of OSPF LSAs
        are listed below in Table 9.

        OSPF routing packets (with the exception of Hellos) are sent
        only over adjacencies.  This means that all OSPF protocol
        packets travel a single IP hop, except those that are sent over
        virtual adjacencies.  The IP source address of an OSPF protocol
        packet is one end of a router adjacency, and the IP destination
        address is either the other end of the adjacency or an IP
        multicast address.


    4.4.  Basic implementation requirements



        An implementation of OSPF requires the following pieces of
        system support:


        Timers
            Two different kind of timers are required.  The first kind,
            called "single shot timers", fire once and cause a protocol
            event to be processed.  The second kind, called "interval
            timers", fire at continuous intervals.  These are used for
            the sending of packets at regular intervals.  A good example
            of this is the regular broadcast of Hello packets. The
            granularity of both kinds of timers is one second.

            Interval timers should be implemented to avoid drift.  In
            some router implementations, packet processing can affect
            timer execution.  When multiple routers are attached to a
            single network, all doing broadcasts, this can lead to the
            synchronization of routing packets (which should be
            avoided).  If timers cannot be implemented to avoid drift,
            small random amounts should be added to/subtracted from the
            interval timer at each firing.





Moy                         Standards Track                    [Page 43]

RFC 2328                     OSPF Version 2                   April 1998




        LS     LSA                LSA description
        type   name
        ________________________________________________________
        1      Router-LSAs        Originated by all routers.
                                  This LSA describes
                                  the collected states of the
                                  router's interfaces to an
                                  area. Flooded throughout a
                                  single area only.
        ________________________________________________________
        2      Network-LSAs       Originated for broadcast
                                  and NBMA networks by
                                  the Designated Router. This
                                  LSA contains the
                                  list of routers connected
                                  to the network. Flooded
                                  throughout a single area only.
        ________________________________________________________
        3,4    Summary-LSAs       Originated by area border
                                  routers, and flooded through-
                                  out the LSA's associated
                                  area. Each summary-LSA
                                  describes a route to a
                                  destination outside the area,
                                  yet still inside the AS
                                  (i.e., an inter-area route).
                                  Type 3 summary-LSAs describe
                                  routes to networks. Type 4
                                  summary-LSAs describe
                                  routes to AS boundary routers.
        ________________________________________________________
        5      AS-external-LSAs   Originated by AS boundary
                                  routers, and flooded through-
                                  out the AS. Each
                                  AS-external-LSA describes
                                  a route to a destination in
                                  another Autonomous System.
                                  Default routes for the AS can
                                  also be described by
                                  AS-external-LSAs.



Moy                         Standards Track                    [Page 44]

RFC 2328                     OSPF Version 2                   April 1998


            Table 9: OSPF link state advertisements (LSAs).



        IP multicast
            Certain OSPF packets take the form of IP multicast
            datagrams.  Support for receiving and sending IP multicast
            datagrams, along with the appropriate lower-level protocol
            support, is required.  The IP multicast datagrams used by
            OSPF never travel more than one hop. For this reason, the
            ability to forward IP multicast datagrams is not required.
            For information on IP multicast, see [Ref7].

        Variable-length subnet support
            The router's IP protocol support must include the ability to
            divide a single IP class A, B, or C network number into many
            subnets of various sizes.  This is commonly called
            variable-length subnetting; see Section 3.5 for details.

        IP supernetting support
            The router's IP protocol support must include the ability to
            aggregate contiguous collections of IP class A, B, and C
            networks into larger quantities called supernets.
            Supernetting has been proposed as one way to improve the
            scaling of IP routing in the worldwide Internet. For more
            information on IP supernetting, see [Ref10].

        Lower-level protocol support
            The lower level protocols referred to here are the network
            access protocols, such as the Ethernet data link layer.
            Indications must be passed from these protocols to OSPF as
            the network interface goes up and down.  For example, on an
            ethernet it would be valuable to know when the ethernet
            transceiver cable becomes unplugged.

        Non-broadcast lower-level protocol support
            On non-broadcast networks, the OSPF Hello Protocol can be
            aided by providing an indication when an attempt is made to
            send a packet to a dead or non-existent router.  For
            example, on an X.25 PDN a dead neighboring router may be





Moy                         Standards Track                    [Page 45]

RFC 2328                     OSPF Version 2                   April 1998


            indicated by the reception of a X.25 clear with an
            appropriate cause and diagnostic, and this information would
            be passed to OSPF.

        List manipulation primitives
            Much of the OSPF functionality is described in terms of its
            operation on lists of LSAs.  For example, the collection of
            LSAs that will be retransmitted to an adjacent router until
            acknowledged are described as a list.  Any particular LSA
            may be on many such lists.  An OSPF implementation needs to
            be able to manipulate these lists, adding and deleting
            constituent LSAs as necessary.

        Tasking support
            Certain procedures described in this specification invoke
            other procedures.  At times, these other procedures should
            be executed in-line, that is, before the current procedure
            is finished.  This is indicated in the text by instructions
            to execute a procedure.  At other times, the other
            procedures are to be executed only when the current
            procedure has finished.  This is indicated by instructions
            to schedule a task.


    4.5.  Optional OSPF capabilities



        The OSPF protocol defines several optional capabilities.  A
        router indicates the optional capabilities that it supports in
        its OSPF Hello packets, Database Description packets and in its
        LSAs.  This enables routers supporting a mix of optional
        capabilities to coexist in a single Autonomous System.

        Some capabilities must be supported by all routers attached to a
        specific area.  In this case, a router will not accept a
        neighbor's Hello Packet unless there is a match in reported
        capabilities (i.e., a capability mismatch prevents a neighbor
        relationship from forming).  An example of this is the
        ExternalRoutingCapability (see below).

        Other capabilities can be negotiated during the Database
        Exchange process.  This is accomplished by specifying the
        optional capabilities in Database Description packets.  A



Moy                         Standards Track                    [Page 46]

RFC 2328                     OSPF Version 2                   April 1998


        capability mismatch with a neighbor in this case will result in
        only a subset of the link state database being exchanged between
        the two neighbors.

        The routing table build process can also be affected by the
        presence/absence of optional capabilities.  For example, since
        the optional capabilities are reported in LSAs, routers
        incapable of certain functions can be avoided when building the
        shortest path tree.

        The OSPF optional capabilities defined in this memo are listed
        below.  See Section A.2 for more information.


        ExternalRoutingCapability
            Entire OSPF areas can be configured as "stubs" (see Section
            3.6).  AS-external-LSAs will not be flooded into stub areas.
            This capability is represented by the E-bit in the OSPF
            Options field (see Section A.2).  In order to ensure
            consistent configuration of stub areas, all routers
            interfacing to such an area must have the E-bit clear in
            their Hello packets (see Sections 9.5 and 10.5).


5.  Protocol Data Structures



    The OSPF protocol is described herein in terms of its operation on
    various protocol data structures.  The following list comprises the
    top-level OSPF data structures.  Any initialization that needs to be
    done is noted.  OSPF areas, interfaces and neighbors also have
    associated data structures that are described later in this
    specification.

    Router ID
        A 32-bit number that uniquely identifies this router in the AS.
        One possible implementation strategy would be to use the
        smallest IP interface address belonging to the router. If a
        router's OSPF Router ID is changed, the router's OSPF software
        should be restarted before the new Router ID takes effect.  In
        this case the router should flush its self-originated LSAs from
        the routing domain (see Section 14.1) before restarting, or they
        will persist for up to MaxAge minutes.



Moy                         Standards Track                    [Page 47]

RFC 2328                     OSPF Version 2                   April 1998


    Area structures
        Each one of the areas to which the router is connected has its
        own data structure.  This data structure describes the working
        of the basic OSPF algorithm.  Remember that each area runs a
        separate copy of the basic OSPF algorithm.

    Backbone (area) structure
        The OSPF backbone area is responsible for the dissemination of
        inter-area routing information.

    Virtual links configured
        The virtual links configured with this router as one endpoint.
        In order to have configured virtual links, the router itself
        must be an area border router.  Virtual links are identified by
        the Router ID of the other endpoint -- which is another area
        border router.  These two endpoint routers must be attached to a
        common area, called the virtual link's Transit area.  Virtual
        links are part of the backbone, and behave as if they were
        unnumbered point-to-point networks between the two routers.  A
        virtual link uses the intra-area routing of its Transit area to
        forward packets.  Virtual links are brought up and down through
        the building of the shortest-path trees for the Transit area.

    List of external routes
        These are routes to destinations external to the Autonomous
        System, that have been gained either through direct experience
        with another routing protocol (such as BGP), or through
        configuration information, or through a combination of the two
        (e.g., dynamic external information to be advertised by OSPF
        with configured metric). Any router having these external routes
        is called an AS boundary router.  These routes are advertised by
        the router into the OSPF routing domain via AS-external-LSAs.

    List of AS-external-LSAs
        Part of the link-state database.  These have originated from the
        AS boundary routers.  They comprise routes to destinations
        external to the Autonomous System.  Note that, if the router is
        itself an AS boundary router, some of these AS-external-LSAs
        have been self-originated.






Moy                         Standards Track                    [Page 48]

RFC 2328                     OSPF Version 2                   April 1998


    The routing table
        Derived from the link-state database.  Each entry in the routing
        table is indexed by a destination, and contains the
        destination's cost and a set of paths to use in forwarding
        packets to the destination. A path is described by its type and
        next hop.  For more information, see Section 11.

    Figure 9 shows the collection of data structures present in a
    typical router.  The router pictured is RT10, from the map in Figure
    6.  Note that Router RT10 has a virtual link configured to Router
    RT11, with Area 2 as the link's Transit area.  This is indicated by
    the dashed line in Figure 9.  When the virtual link becomes active,
    through the building of the shortest path tree for Area 2, it
    becomes an interface to the backbone (see the two backbone
    interfaces depicted in Figure 9).

6.  The Area Data Structure



    The area data structure contains all the information used to run the
    basic OSPF routing algorithm. Each area maintains its own link-state
    database. A network belongs to a single area, and a router interface
    connects to a single area. Each router adjacency also belongs to a
    single area.

    The OSPF backbone is the special OSPF area responsible for
    disseminating inter-area routing information.

    The area link-state database consists of the collection of router-
    LSAs, network-LSAs and summary-LSAs that have originated from the
    area's routers.  This information is flooded throughout a single
    area only.  The list of AS-external-LSAs (see Section 5) is also
    considered to be part of each area's link-state database.

    Area ID
        A 32-bit number identifying the area. The Area ID of 0.0.0.0 is
        reserved for the backbone.

    List of area address ranges
        In order to aggregate routing information at area boundaries,
        area address ranges can be employed. Each address range is
        specified by an [address,mask] pair and a status indication of
        either Advertise or DoNotAdvertise (see Section 12.4.3).



Moy                         Standards Track                    [Page 49]

RFC 2328                     OSPF Version 2                   April 1998





                              +----+
                              |RT10|------+
                              +----+       \+-------------+
                             /      \       |Routing Table|
                            /        \      +-------------+
                           /          \
              +------+    /            \    +--------+
              |Area 2|---+              +---|Backbone|
              +------+***********+          +--------+
             /        \           *        /          \
            /          \           *      /            \
       +---------+  +---------+    +------------+       +------------+
       |Interface|  |Interface|    |Virtual Link|       |Interface Ib|
       |  to N6  |  |  to N8  |    |   to RT11  |       +------------+
       +---------+  +---------+    +------------+             |
           /  \           |               |                   |
          /    \          |               |                   |
   +--------+ +--------+  |        +-------------+      +------------+
   |Neighbor| |Neighbor|  |        |Neighbor RT11|      |Neighbor RT6|
   |  RT8   | |  RT7   |  |        +-------------+      +------------+
   +--------+ +--------+  |
                          |
                     +-------------+
                     |Neighbor RT11|
                     +-------------+


                Figure 9: Router RT10's Data structures

    Associated router interfaces
        This router's interfaces connecting to the area.  A router
        interface belongs to one and only one area (or the backbone).
        For the backbone area this list includes all the virtual links.
        A virtual link is identified by the Router ID of its other
        endpoint; its cost is the cost of the shortest intra-area path
        through the Transit area that exists between the two routers.






Moy                         Standards Track                    [Page 50]

RFC 2328                     OSPF Version 2                   April 1998


    List of router-LSAs
        A router-LSA is generated by each router in the area.  It
        describes the state of the router's interfaces to the area.

    List of network-LSAs
        One network-LSA is generated for each transit broadcast and NBMA
        network in the area.  A network-LSA describes the set of routers
        currently connected to the network.

    List of summary-LSAs
        Summary-LSAs originate from the area's area border routers.
        They describe routes to destinations internal to the Autonomous
        System, yet external to the area (i.e., inter-area
        destinations).

    Shortest-path tree
        The shortest-path tree for the area, with this router itself as
        root.  Derived from the collected router-LSAs and network-LSAs
        by the Dijkstra algorithm (see Section 16.1).

    TransitCapability
        This parameter indicates whether the area can carry data traffic
        that neither originates nor terminates in the area itself. This
        parameter is calculated when the area's shortest-path tree is
        built (see Section 16.1, where TransitCapability is set to TRUE
        if and only if there are one or more fully adjacent virtual
        links using the area as Transit area), and is used as an input
        to a subsequent step of the routing table build process (see
        Section 16.3). When an area's TransitCapability is set to TRUE,
        the area is said to be a "transit area".

    ExternalRoutingCapability
        Whether AS-external-LSAs will be flooded into/throughout the
        area.  This is a configurable parameter.  If AS-external-LSAs
        are excluded from the area, the area is called a "stub". Within
        stub areas, routing to AS external destinations will be based
        solely on a default summary route.  The backbone cannot be
        configured as a stub area.  Also, virtual links cannot be
        configured through stub areas.  For more information, see
        Section 3.6.





Moy                         Standards Track                    [Page 51]

RFC 2328                     OSPF Version 2                   April 1998


    StubDefaultCost
        If the area has been configured as a stub area, and the router
        itself is an area border router, then the StubDefaultCost
        indicates the cost of the default summary-LSA that the router
        should advertise into the area. See Section 12.4.3 for more
        information.


    Unless otherwise specified, the remaining sections of this document
    refer to the operation of the OSPF protocol within a single area.


7.  Bringing Up Adjacencies



    OSPF creates adjacencies between neighboring routers for the purpose
    of exchanging routing information.  Not every two neighboring
    routers will become adjacent.  This section covers the generalities
    involved in creating adjacencies.  For further details consult
    Section 10.


    7.1.  The Hello Protocol



        The Hello Protocol is responsible for establishing and
        maintaining neighbor relationships.  It also ensures that
        communication between neighbors is bidirectional.  Hello packets
        are sent periodically out all router interfaces.  Bidirectional
        communication is indicated when the router sees itself listed in
        the neighbor's Hello Packet.  On broadcast and NBMA networks,
        the Hello Protocol elects a Designated Router for the network.

        The Hello Protocol works differently on broadcast networks, NBMA
        networks and Point-to-MultiPoint networks.  On broadcast
        networks, each router advertises itself by periodically
        multicasting Hello Packets.  This allows neighbors to be
        discovered dynamically.  These Hello Packets contain the
        router's view of the Designated Router's identity, and the list
        of routers whose Hello Packets have been seen recently.

        On NBMA networks some configuration information may be necessary
        for the operation of the Hello Protocol.  Each router that may
        potentially become Designated Router has a list of all other



Moy                         Standards Track                    [Page 52]

RFC 2328                     OSPF Version 2                   April 1998


        routers attached to the network.  A router, having Designated
        Router potential, sends Hello Packets to all other potential
        Designated Routers when its interface to the NBMA network first
        becomes operational.  This is an attempt to find the Designated
        Router for the network.  If the router itself is elected
        Designated Router, it begins sending Hello Packets to all other
        routers attached to the network.

        On Point-to-MultiPoint networks, a router sends Hello Packets to
        all neighbors with which it can communicate directly. These
        neighbors may be discovered dynamically through a protocol such
        as Inverse ARP (see [Ref14]), or they may be configured.

        After a neighbor has been discovered, bidirectional
        communication ensured, and (if on a broadcast or NBMA network) a
        Designated Router elected, a decision is made regarding whether
        or not an adjacency should be formed with the neighbor (see
        Section 10.4). If an adjacency is to be formed, the first step
        is to synchronize the neighbors' link-state databases.  This is
        covered in the next section.


    7.2.  The Synchronization of Databases



        In a link-state routing algorithm, it is very important for all
        routers' link-state databases to stay synchronized.  OSPF
        simplifies this by requiring only adjacent routers to remain
        synchronized.  The synchronization process begins as soon as the
        routers attempt to bring up the adjacency.  Each router
        describes its database by sending a sequence of Database
        Description packets to its neighbor.  Each Database Description
        Packet describes a set of LSAs belonging to the router's
        database.  When the neighbor sees an LSA that is more recent
        than its own database copy, it makes a note that this newer LSA
        should be requested.

        This sending and receiving of Database Description packets is
        called the "Database Exchange Process".  During this process,
        the two routers form a master/slave relationship.  Each Database
        Description Packet has a sequence number.  Database Description
        Packets sent by the master (polls) are acknowledged by the slave
        through echoing of the sequence number.  Both polls and their



Moy                         Standards Track                    [Page 53]

RFC 2328                     OSPF Version 2                   April 1998


        responses contain summaries of link state data.  The master is
        the only one allowed to retransmit Database Description Packets.
        It does so only at fixed intervals, the length of which is the
        configured per-interface constant RxmtInterval.

        Each Database Description contains an indication that there are
        more packets to follow --- the M-bit.  The Database Exchange
        Process is over when a router has received and sent Database
        Description Packets with the M-bit off.

        During and after the Database Exchange Process, each router has
        a list of those LSAs for which the neighbor has more up-to-date
        instances.  These LSAs are requested in Link State Request
        Packets.  Link State Request packets that are not satisfied are
        retransmitted at fixed intervals of time RxmtInterval.  When the
        Database Description Process has completed and all Link State
        Requests have been satisfied, the databases are deemed
        synchronized and the routers are marked fully adjacent.  At this
        time the adjacency is fully functional and is advertised in the
        two routers' router-LSAs.

        The adjacency is used by the flooding procedure as soon as the
        Database Exchange Process begins.  This simplifies database
        synchronization, and guarantees that it finishes in a
        predictable period of time.


    7.3.  The Designated Router



        Every broadcast and NBMA network has a Designated Router.  The
        Designated Router performs two main functions for the routing
        protocol:

        o   The Designated Router originates a network-LSA on behalf of
            the network.  This LSA lists the set of routers (including
            the Designated Router itself) currently attached to the
            network.  The Link State ID for this LSA (see Section
            12.1.4) is the IP interface address of the Designated
            Router.  The IP network number can then be obtained by using
            the network's subnet/network mask.





Moy                         Standards Track                    [Page 54]

RFC 2328                     OSPF Version 2                   April 1998


        o   The Designated Router becomes adjacent to all other routers
            on the network.  Since the link state databases are
            synchronized across adjacencies (through adjacency bring-up
            and then the flooding procedure), the Designated Router
            plays a central part in the synchronization process.


        The Designated Router is elected by the Hello Protocol.  A
        router's Hello Packet contains its Router Priority, which is
        configurable on a per-interface basis.  In general, when a
        router's interface to a network first becomes functional, it
        checks to see whether there is currently a Designated Router for
        the network.  If there is, it accepts that Designated Router,
        regardless of its Router Priority.  (This makes it harder to
        predict the identity of the Designated Router, but ensures that
        the Designated Router changes less often.  See below.)
        Otherwise, the router itself becomes Designated Router if it has
        the highest Router Priority on the network.  A more detailed
        (and more accurate) description of Designated Router election is
        presented in Section 9.4.

        The Designated Router is the endpoint of many adjacencies.  In
        order to optimize the flooding procedure on broadcast networks,
        the Designated Router multicasts its Link State Update Packets
        to the address AllSPFRouters, rather than sending separate
        packets over each adjacency.

        Section 2 of this document discusses the directed graph
        representation of an area.  Router nodes are labelled with their
        Router ID.  Transit network nodes are actually labelled with the
        IP address of their Designated Router.  It follows that when the
        Designated Router changes, it appears as if the network node on
        the graph is replaced by an entirely new node.  This will cause
        the network and all its attached routers to originate new LSAs.
        Until the link-state databases again converge, some temporary
        loss of connectivity may result.  This may result in ICMP
        unreachable messages being sent in response to data traffic.
        For that reason, the Designated Router should change only
        infrequently.  Router Priorities should be configured so that
        the most dependable router on a network eventually becomes
        Designated Router.




Moy                         Standards Track                    [Page 55]

RFC 2328                     OSPF Version 2                   April 1998


    7.4.  The Backup Designated Router



        In order to make the transition to a new Designated Router
        smoother, there is a Backup Designated Router for each broadcast
        and NBMA network.  The Backup Designated Router is also adjacent
        to all routers on the network, and becomes Designated Router
        when the previous Designated Router fails.  If there were no
        Backup Designated Router, when a new Designated Router became
        necessary, new adjacencies would have to be formed between the
        new Designated Router and all other routers attached to the
        network.  Part of the adjacency forming process is the
        synchronizing of link-state databases, which can potentially
        take quite a long time.  During this time, the network would not
        be available for transit data traffic.  The Backup Designated
        obviates the need to form these adjacencies, since they already
        exist.  This means the period of disruption in transit traffic
        lasts only as long as it takes to flood the new LSAs (which
        announce the new Designated Router).

        The Backup Designated Router does not generate a network-LSA for
        the network.  (If it did, the transition to a new Designated
        Router would be even faster.  However, this is a tradeoff
        between database size and speed of convergence when the
        Designated Router disappears.)

        The Backup Designated Router is also elected by the Hello
        Protocol.  Each Hello Packet has a field that specifies the
        Backup Designated Router for the network.

        In some steps of the flooding procedure, the Backup Designated
        Router plays a passive role, letting the Designated Router do
        more of the work.  This cuts down on the amount of local routing
        traffic.  See Section 13.3 for more information.


    7.5.  The graph of adjacencies



        An adjacency is bound to the network that the two routers have
        in common.  If two routers have multiple networks in common,
        they may have multiple adjacencies between them.





Moy                         Standards Track                    [Page 56]

RFC 2328                     OSPF Version 2                   April 1998


        One can picture the collection of adjacencies on a network as
        forming an undirected graph.  The vertices consist of routers,
        with an edge joining two routers if they are adjacent.  The
        graph of adjacencies describes the flow of routing protocol
        packets, and in particular Link State Update Packets, through
        the Autonomous System.

        Two graphs are possible, depending on whether a Designated
        Router is elected for the network.  On physical point-to-point
        networks, Point-to-MultiPoint networks and virtual links,
        neighboring routers become adjacent whenever they can
        communicate directly.  In contrast, on broadcast and NBMA
        networks only the Designated Router and the Backup Designated
        Router become adjacent to all other routers attached to the
        network.



          +---+            +---+
          |RT1|------------|RT2|            o---------------o
          +---+    N1      +---+           RT1             RT2



                                                 RT7
                                                  o---------+
            +---+   +---+   +---+                /|\        |
            |RT7|   |RT3|   |RT4|               / | \       |
            +---+   +---+   +---+              /  |  \      |
              |       |       |               /   |   \     |
         +-----------------------+        RT5o RT6o    oRT4 |
                  |       |     N2            *   *   *     |
                +---+   +---+                  *  *  *      |
                |RT5|   |RT6|                   * * *       |
                +---+   +---+                    ***        |
                                                  o---------+
                                                 RT3


                  Figure 10: The graph of adjacencies





Moy                         Standards Track                    [Page 57]

RFC 2328                     OSPF Version 2                   April 1998


        These graphs are shown in Figure 10.  It is assumed that Router
        RT7 has become the Designated Router, and Router RT3 the Backup
        Designated Router, for the Network N2.  The Backup Designated
        Router performs a lesser function during the flooding procedure
        than the Designated Router (see Section 13.3).  This is the
        reason for the dashed lines connecting the Backup Designated
        Router RT3.


8.  Protocol Packet Processing



    This section discusses the general processing of OSPF routing
    protocol packets.  It is very important that the router link-state
    databases remain synchronized.  For this reason, routing protocol
    packets should get preferential treatment over ordinary data
    packets, both in sending and receiving.

    Routing protocol packets are sent along adjacencies only (with the
    exception of Hello packets, which are used to discover the
    adjacencies).  This means that all routing protocol packets travel a
    single IP hop, except those sent over virtual links.

    All routing protocol packets begin with a standard header.  The
    sections below provide details on how to fill in and verify this
    standard header.  Then, for each packet type, the section giving
    more details on that particular packet type's processing is listed.

    8.1.  Sending protocol packets



        When a router sends a routing protocol packet, it fills in the
        fields of the standard OSPF packet header as follows.  For more
        details on the header format consult Section A.3.1:

        Version #
            Set to 2, the version number of the protocol as documented
            in this specification.

        Packet type
            The type of OSPF packet, such as Link state Update or Hello
            Packet.





Moy                         Standards Track                    [Page 58]

RFC 2328                     OSPF Version 2                   April 1998


        Packet length
            The length of the entire OSPF packet in bytes, including the
            standard OSPF packet header.

        Router ID
            The identity of the router itself (who is originating the
            packet).

        Area ID
            The OSPF area that the packet is being sent into.

        Checksum
            The standard IP 16-bit one's complement checksum of the
            entire OSPF packet, excluding the 64-bit authentication
            field.  This checksum is calculated as part of the
            appropriate authentication procedure; for some OSPF
            authentication types, the checksum calculation is omitted.
            See Section D.4 for details.

        AuType and Authentication
            Each OSPF packet exchange is authenticated.  Authentication
            types are assigned by the protocol and are documented in
            Appendix D.  A different authentication procedure can be
            used for each IP network/subnet.  Autype indicates the type
            of authentication procedure in use. The 64-bit
            authentication field is then for use by the chosen
            authentication procedure.  This procedure should be the last
            called when forming the packet to be sent. See Section D.4
            for details.


        The IP destination address for the packet is selected as
        follows.  On physical point-to-point networks, the IP
        destination is always set to the address AllSPFRouters.  On all
        other network types (including virtual links), the majority of
        OSPF packets are sent as unicasts, i.e., sent directly to the
        other end of the adjacency.  In this case, the IP destination is
        just the Neighbor IP address associated with the other end of
        the adjacency (see Section 10).  The only packets not sent as
        unicasts are on broadcast networks; on these networks Hello
        packets are sent to the multicast destination AllSPFRouters, the
        Designated Router and its Backup send both Link State Update



Moy                         Standards Track                    [Page 59]

RFC 2328                     OSPF Version 2                   April 1998


        Packets and Link State Acknowledgment Packets to the multicast
        address AllSPFRouters, while all other routers send both their
        Link State Update and Link State Acknowledgment Packets to the
        multicast address AllDRouters.

        Retransmissions of Link State Update packets are ALWAYS sent
        directly to the neighbor. On multi-access networks, this means
        that retransmissions should be sent to the neighbor's IP
        address.

        The IP source address should be set to the IP address of the
        sending interface.  Interfaces to unnumbered point-to-point
        networks have no associated IP address.  On these interfaces,
        the IP source should be set to any of the other IP addresses
        belonging to the router.  For this reason, there must be at
        least one IP address assigned to the router.[2] Note that, for
        most purposes, virtual links act precisely the same as
        unnumbered point-to-point networks.  However, each virtual link
        does have an IP interface address (discovered during the routing
        table build process) which is used as the IP source when sending
        packets over the virtual link.

        For more information on the format of specific OSPF packet
        types, consult the sections listed in Table 10.



             Type   Packet name            detailed section (transmit)
             _________________________________________________________
             1      Hello                  Section  9.5
             2      Database description   Section 10.8
             3      Link state request     Section 10.9
             4      Link state update      Section 13.3
             5      Link state ack         Section 13.5


      Table 10: Sections describing OSPF protocol packet transmission.








Moy                         Standards Track                    [Page 60]

RFC 2328                     OSPF Version 2                   April 1998


    8.2.  Receiving protocol packets



        Whenever a protocol packet is received by the router it is
        marked with the interface it was received on.  For routers that
        have virtual links configured, it may not be immediately obvious
        which interface to associate the packet with.  For example,
        consider the Router RT11 depicted in Figure 6.  If RT11 receives
        an OSPF protocol packet on its interface to Network N8, it may
        want to associate the packet with the interface to Area 2, or
        with the virtual link to Router RT10 (which is part of the
        backbone).  In the following, we assume that the packet is
        initially associated with the non-virtual  link.[3]

        In order for the packet to be accepted at the IP level, it must
        pass a number of tests, even before the packet is passed to OSPF
        for processing:


        o   The IP checksum must be correct.

        o   The packet's IP destination address must be the IP address
            of the receiving interface, or one of the IP multicast
            addresses AllSPFRouters or AllDRouters.

        o   The IP protocol specified must be OSPF (89).

        o   Locally originated packets should not be passed on to OSPF.
            That is, the source IP address should be examined to make
            sure this is not a multicast packet that the router itself
            generated.


        Next, the OSPF packet header is verified.  The fields specified
        in the header must match those configured for the receiving
        interface.  If they do not, the packet should be discarded:


        o   The version number field must specify protocol version 2.

        o   The Area ID found in the OSPF header must be verified.  If
            both of the following cases fail, the packet should be
            discarded.  The Area ID specified in the header must either:



Moy                         Standards Track                    [Page 61]

RFC 2328                     OSPF Version 2                   April 1998


            (1) Match the Area ID of the receiving interface.  In this
                case, the packet has been sent over a single hop.
                Therefore, the packet's IP source address is required to
                be on the same network as the receiving interface.  This
                can be verified by comparing the packet's IP source
                address to the interface's IP address, after masking
                both addresses with the interface mask.  This comparison
                should not be performed on point-to-point networks. On
                point-to-point networks, the interface addresses of each
                end of the link are assigned independently, if they are
                assigned at all.

            (2) Indicate the backbone.  In this case, the packet has
                been sent over a virtual link.  The receiving router
                must be an area border router, and the Router ID
                specified in the packet (the source router) must be the
                other end of a configured virtual link.  The receiving
                interface must also attach to the virtual link's
                configured Transit area.  If all of these checks
                succeed, the packet is accepted and is from now on
                associated with the virtual link (and the backbone
                area).

        o   Packets whose IP destination is AllDRouters should only be
            accepted if the state of the receiving interface is DR or
            Backup (see Section 9.1).

        o   The AuType specified in the packet must match the AuType
            specified for the associated area.

        o   The packet must be authenticated.  The authentication
            procedure is indicated by the setting of AuType (see
            Appendix D).  The authentication procedure may use one or
            more Authentication keys, which can be configured on a per-
            interface basis.  The authentication procedure may also
            verify the checksum field in the OSPF packet header (which,
            when used, is set to the standard IP 16-bit one's complement
            checksum of the OSPF packet's contents after excluding the
            64-bit authentication field).  If the authentication
            procedure fails, the packet should be discarded.





Moy                         Standards Track                    [Page 62]

RFC 2328                     OSPF Version 2                   April 1998


        If the packet type is Hello, it should then be further processed
        by the Hello Protocol (see Section 10.5).  All other packet
        types are sent/received only on adjacencies.  This means that
        the packet must have been sent by one of the router's active
        neighbors.  If the receiving interface connects to a broadcast
        network, Point-to-MultiPoint network or NBMA network the sender
        is identified by the IP source address found in the packet's IP
        header.  If the receiving interface connects to a point-to-point
        network or a virtual link, the sender is identified by the
        Router ID (source router) found in the packet's OSPF header.
        The data structure associated with the receiving interface
        contains the list of active neighbors.  Packets not matching any
        active neighbor are discarded.

        At this point all received protocol packets are associated with
        an active neighbor.  For the further input processing of
        specific packet types, consult the sections listed in Table 11.



              Type   Packet name            detailed section (receive)
              ________________________________________________________
              1      Hello                  Section 10.5
              2      Database description   Section 10.6
              3      Link state request     Section 10.7
              4      Link state update      Section 13
              5      Link state ack         Section 13.7


      Table 11: Sections describing OSPF protocol packet reception.



9.  The Interface Data Structure



    An OSPF interface is the connection between a router and a network.
    We assume a single OSPF interface to each attached network/subnet,
    although supporting multiple interfaces on a single network is
    considered in Appendix F. Each interface structure has at most one
    IP interface address.





Moy                         Standards Track                    [Page 63]

RFC 2328                     OSPF Version 2                   April 1998


    An OSPF interface can be considered to belong to the area that
    contains the attached network.  All routing protocol packets
    originated by the router over this interface are labelled with the
    interface's Area ID.  One or more router adjacencies may develop
    over an interface.  A router's LSAs reflect the state of its
    interfaces and their associated adjacencies.

    The following data items are associated with an interface.  Note
    that a number of these items are actually configuration for the
    attached network; such items must be the same for all routers
    connected to the network.

    Type
        The OSPF interface type is either point-to-point, broadcast,
        NBMA, Point-to-MultiPoint or virtual link.

    State
        The functional level of an interface.  State determines whether
        or not full adjacencies are allowed to form over the interface.
        State is also reflected in the router's LSAs.

    IP interface address
        The IP address associated with the interface.  This appears as
        the IP source address in all routing protocol packets originated
        over this interface.  Interfaces to unnumbered point-to-point
        networks do not have an associated IP address.

    IP interface mask
        Also referred to as the subnet mask, this indicates the portion
        of the IP interface address that identifies the attached
        network.  Masking the IP interface address with the IP interface
        mask yields the IP network number of the attached network.  On
        point-to-point networks and virtual links, the IP interface mask
        is not defined. On these networks, the link itself is not
        assigned an IP network number, and so the addresses of each side
        of the link are assigned independently, if they are assigned at
        all.

    Area ID
        The Area ID of the area to which the attached network belongs.
        All routing protocol packets originating from the interface are
        labelled with this Area ID.



Moy                         Standards Track                    [Page 64]

RFC 2328                     OSPF Version 2                   April 1998


    HelloInterval
        The length of time, in seconds, between the Hello packets that
        the router sends on the interface.  Advertised in Hello packets
        sent out this interface.

    RouterDeadInterval
        The number of seconds before the router's neighbors will declare
        it down, when they stop hearing the router's Hello Packets.
        Advertised in Hello packets sent out this interface.

    InfTransDelay
        The estimated number of seconds it takes to transmit a Link
        State Update Packet over this interface.  LSAs contained in the
        Link State Update packet will have their age incremented by this
        amount before transmission.  This value should take into account
        transmission and propagation delays; it must be greater than
        zero.

    Router Priority
        An 8-bit unsigned integer.  When two routers attached to a
        network both attempt to become Designated Router, the one with
        the highest Router Priority takes precedence.  A router whose
        Router Priority is set to 0 is ineligible to become Designated
        Router on the attached network.  Advertised in Hello packets
        sent out this interface.

    Hello Timer
        An interval timer that causes the interface to send a Hello
        packet.  This timer fires every HelloInterval seconds.  Note
        that on non-broadcast networks a separate Hello packet is sent
        to each qualified neighbor.

    Wait Timer
        A single shot timer that causes the interface to exit the
        Waiting state, and as a consequence select a Designated Router
        on the network.  The length of the timer is RouterDeadInterval
        seconds.

    List of neighboring routers
        The other routers attached to this network.  This list is formed
        by the Hello Protocol.  Adjacencies will be formed to some of




Moy                         Standards Track                    [Page 65]

RFC 2328                     OSPF Version 2                   April 1998


        these neighbors.  The set of adjacent neighbors can be
        determined by an examination of all of the neighbors' states.

    Designated Router
        The Designated Router selected for the attached network.  The
        Designated Router is selected on all broadcast and NBMA networks
        by the Hello Protocol.  Two pieces of identification are kept
        for the Designated Router: its Router ID and its IP interface
        address on the network.  The Designated Router advertises link
        state for the network; this network-LSA is labelled with the
        Designated Router's IP address.  The Designated Router is
        initialized to 0.0.0.0, which indicates the lack of a Designated
        Router.

    Backup Designated Router
        The Backup Designated Router is also selected on all broadcast
        and NBMA networks by the Hello Protocol.  All routers on the
        attached network become adjacent to both the Designated Router
        and the Backup Designated Router.  The Backup Designated Router
        becomes Designated Router when the current Designated Router
        fails.  The Backup Designated Router is initialized to 0.0.0.0,
        indicating the lack of a Backup Designated Router.

    Interface output cost(s)
        The cost of sending a data packet on the interface, expressed in
        the link state metric.  This is advertised as the link cost for
        this interface in the router-LSA. The cost of an interface must
        be greater than zero.

    RxmtInterval
        The number of seconds between LSA retransmissions, for
        adjacencies belonging to this interface.  Also used when
        retransmitting Database Description and Link State Request
        Packets.

    AuType
        The type of authentication used on the attached network/subnet.
        Authentication types are defined in Appendix D.  All OSPF packet
        exchanges are authenticated.  Different authentication schemes
        may be used on different networks/subnets.





Moy                         Standards Track                    [Page 66]

RFC 2328                     OSPF Version 2                   April 1998


    Authentication key
        This configured data allows the authentication procedure to
        generate and/or verify OSPF protocol packets.  The
        Authentication key can be configured on a per-interface basis.
        For example, if the AuType indicates simple password, the
        Authentication key would be a 64-bit clear password which is
        inserted into the OSPF packet header. If instead Autype
        indicates Cryptographic authentication, then the Authentication
        key is a shared secret which enables the generation/verification
        of message digests which are appended to the OSPF protocol
        packets. When Cryptographic authentication is used, multiple
        simultaneous keys are supported in order to achieve smooth key
        transition (see Section D.3).


    9.1.  Interface states



        The various states that router interfaces may attain is
        documented in this section.  The states are listed in order of
        progressing functionality.  For example, the inoperative state
        is listed first, followed by a list of intermediate states
        before the final, fully functional state is achieved.  The
        specification makes use of this ordering by sometimes making
        references such as "those interfaces in state greater than X".
        Figure 11 shows the graph of interface state changes.  The arcs
        of the graph are labelled with the event causing the state
        change.  These events are documented in Section 9.2.  The
        interface state machine is described in more detail in Section
        9.3.


        Down
            This is the initial interface state.  In this state, the
            lower-level protocols have indicated that the interface is
            unusable.  No protocol traffic at all will be sent or
            received on such a interface.  In this state, interface
            parameters should be set to their initial values.  All
            interface timers should be disabled, and there should be no
            adjacencies associated with the interface.

        Loopback
            In this state, the router's interface to the network is



Moy                         Standards Track                    [Page 67]

RFC 2328                     OSPF Version 2                   April 1998



                                  +----+   UnloopInd   +--------+
                                  |Down|<--------------|Loopback|
                                  +----+               +--------+
                                     |
                                     |InterfaceUp
                          +-------+  |               +--------------+
                          |Waiting|<-+-------------->|Point-to-point|
                          +-------+                  +--------------+
                              |
                     WaitTimer|BackupSeen
                              |
                              |
                              |   NeighborChange
          +------+           +-+<---------------- +-------+
          |Backup|<----------|?|----------------->|DROther|
          +------+---------->+-+<-----+           +-------+
                    Neighbor  |       |
                    Change    |       |Neighbor
                              |       |Change
                              |     +--+
                              +---->|DR|
                                    +--+

                      Figure 11: Interface State changes

                 In addition to the state transitions pictured,
                 Event InterfaceDown always forces Down State, and
                 Event LoopInd always forces Loopback State


            looped back.  The interface may be looped back in hardware
            or software.  The interface will be unavailable for regular
            data traffic.  However, it may still be desirable to gain
            information on the quality of this interface, either through
            sending ICMP pings to the interface or through something
            like a bit error test.  For this reason, IP packets may
            still be addressed to an interface in Loopback state.  To







Moy                         Standards Track                    [Page 68]

RFC 2328                     OSPF Version 2                   April 1998


            facilitate this, such interfaces are advertised in router-
            LSAs as single host routes, whose destination is the IP
            interface address.[4]

        Waiting
            In this state, the router is trying to determine the
            identity of the (Backup) Designated Router for the network.
            To do this, the router monitors the Hello Packets it
            receives.  The router is not allowed to elect a Backup
            Designated Router nor a Designated Router until it
            transitions out of Waiting state.  This prevents unnecessary
            changes of (Backup) Designated Router.

        Point-to-point
            In this state, the interface is operational, and connects
            either to a physical point-to-point network or to a virtual
            link.  Upon entering this state, the router attempts to form
            an adjacency with the neighboring router.  Hello Packets are
            sent to the neighbor every HelloInterval seconds.

        DR Other
            The interface is to a broadcast or NBMA network on which
            another router has been selected to be the Designated
            Router.  In this state, the router itself has not been
            selected Backup Designated Router either.  The router forms
            adjacencies to both the Designated Router and the Backup
            Designated Router (if they exist).

        Backup
            In this state, the router itself is the Backup Designated
            Router on the attached network.  It will be promoted to
            Designated Router when the present Designated Router fails.
            The router establishes adjacencies to all other routers
            attached to the network.  The Backup Designated Router
            performs slightly different functions during the Flooding
            Procedure, as compared to the Designated Router (see Section
            13.3).  See Section 7.4 for more details on the functions
            performed by the Backup Designated Router.

        DR  In this state, this router itself is the Designated Router
            on the attached network.  Adjacencies are established to all
            other routers attached to the network.  The router must also



Moy                         Standards Track                    [Page 69]

RFC 2328                     OSPF Version 2                   April 1998


            originate a network-LSA for the network node.  The network-
            LSA will contain links to all routers (including the
            Designated Router itself) attached to the network.  See
            Section 7.3 for more details on the functions performed by
            the Designated Router.


    9.2.  Events causing interface state changes



        State changes can be effected by a number of events.  These
        events are pictured as the labelled arcs in Figure 11.  The
        label definitions are listed below.  For a detailed explanation
        of the effect of these events on OSPF protocol operation,
        consult Section 9.3.


        InterfaceUp
            Lower-level protocols have indicated that the network
            interface is operational.  This enables the interface to
            transition out of Down state.  On virtual links, the
            interface operational indication is actually a result of the
            shortest path calculation (see Section 16.7).

        WaitTimer
            The Wait Timer has fired, indicating the end of the waiting
            period that is required before electing a (Backup)
            Designated Router.

        BackupSeen
            The router has detected the existence or non-existence of a
            Backup Designated Router for the network.  This is done in
            one of two ways.  First, an Hello Packet may be received
            from a neighbor claiming to be itself the Backup Designated
            Router.  Alternatively, an Hello Packet may be received from
            a neighbor claiming to be itself the Designated Router, and
            indicating that there is no Backup Designated Router.  In
            either case there must be bidirectional communication with
            the neighbor, i.e., the router must also appear in the
            neighbor's Hello Packet.  This event signals an end to the
            Waiting state.





Moy                         Standards Track                    [Page 70]

RFC 2328                     OSPF Version 2                   April 1998


        NeighborChange
            There has been a change in the set of bidirectional
            neighbors associated with the interface.  The (Backup)
            Designated Router needs to be recalculated.  The following
            neighbor changes lead to the NeighborChange event.  For an
            explanation of neighbor states, see Section 10.1.

            o   Bidirectional communication has been established to a
                neighbor.  In other words, the state of the neighbor has
                transitioned to 2-Way or higher.

            o   There is no longer bidirectional communication with a
                neighbor.  In other words, the state of the neighbor has
                transitioned to Init or lower.

            o   One of the bidirectional neighbors is newly declaring
                itself as either Designated Router or Backup Designated
                Router.  This is detected through examination of that
                neighbor's Hello Packets.

            o   One of the bidirectional neighbors is no longer
                declaring itself as Designated Router, or is no longer
                declaring itself as Backup Designated Router.  This is
                again detected through examination of that neighbor's
                Hello Packets.

            o   The advertised Router Priority for a bidirectional
                neighbor has changed.  This is again detected through
                examination of that neighbor's Hello Packets.

        LoopInd
            An indication has been received that the interface is now
            looped back to itself.  This indication can be received
            either from network management or from the lower level
            protocols.

        UnloopInd
            An indication has been received that the interface is no
            longer looped back.  As with the LoopInd event, this






Moy                         Standards Track                    [Page 71]

RFC 2328                     OSPF Version 2                   April 1998


            indication can be received either from network management or
            from the lower level protocols.

        InterfaceDown
            Lower-level protocols indicate that this interface is no
            longer functional.  No matter what the current interface
            state is, the new interface state will be Down.

    9.3.  The Interface state machine



        A detailed description of the interface state changes follows.
        Each state change is invoked by an event (Section 9.2).  This
        event may produce different effects, depending on the current
        state of the interface.  For this reason, the state machine
        below is organized by current interface state and received
        event.  Each entry in the state machine describes the resulting
        new interface state and the required set of additional actions.

        When an interface's state changes, it may be necessary to
        originate a new router-LSA.  See Section 12.4 for more details.

        Some of the required actions below involve generating events for
        the neighbor state machine.  For example, when an interface
        becomes inoperative, all neighbor connections associated with
        the interface must be destroyed.  For more information on the
        neighbor state machine, see Section 10.3.


         State(s):  Down

            Event:  InterfaceUp

        New state:  Depends upon action routine

           Action:  Start the interval Hello Timer, enabling the
                    periodic sending of Hello packets out the interface.
                    If the attached network is a physical point-to-point
                    network, Point-to-MultiPoint network or virtual
                    link, the interface state transitions to Point-to-
                    Point.  Else, if the router is not eligible to
                    become Designated Router the interface state
                    transitions to DR Other.



Moy                         Standards Track                    [Page 72]

RFC 2328                     OSPF Version 2                   April 1998


                    Otherwise, the attached network is a broadcast or
                    NBMA network and the router is eligible to become
                    Designated Router.  In this case, in an attempt to
                    discover the attached network's Designated Router
                    the interface state is set to Waiting and the single
                    shot Wait Timer is started.  Additionally, if the
                    network is an NBMA network examine the configured
                    list of neighbors for this interface and generate
                    the neighbor event Start for each neighbor that is
                    also eligible to become Designated Router.


         State(s):  Waiting

            Event:  BackupSeen

        New state:  Depends upon action routine.

           Action:  Calculate the attached network's Backup Designated
                    Router and Designated Router, as shown in Section
                    9.4.  As a result of this calculation, the new state
                    of the interface will be either DR Other, Backup or
                    DR.


         State(s):  Waiting

            Event:  WaitTimer

        New state:  Depends upon action routine.

           Action:  Calculate the attached network's Backup Designated
                    Router and Designated Router, as shown in Section
                    9.4.  As a result of this calculation, the new state
                    of the interface will be either DR Other, Backup or
                    DR.


         State(s):  DR Other, Backup or DR

            Event:  NeighborChange




Moy                         Standards Track                    [Page 73]

RFC 2328                     OSPF Version 2                   April 1998


        New state:  Depends upon action routine.

           Action:  Recalculate the attached network's Backup Designated
                    Router and Designated Router, as shown in Section
                    9.4.  As a result of this calculation, the new state
                    of the interface will be either DR Other, Backup or
                    DR.


         State(s):  Any State

            Event:  InterfaceDown

        New state:  Down

           Action:  All interface variables are reset, and interface
                    timers disabled.  Also, all neighbor connections
                    associated with the interface are destroyed.  This
                    is done by generating the event KillNbr on all
                    associated neighbors (see Section 10.2).


         State(s):  Any State

            Event:  LoopInd

        New state:  Loopback

           Action:  Since this interface is no longer connected to the
                    attached network the actions associated with the
                    above InterfaceDown event are executed.


         State(s):  Loopback

            Event:  UnloopInd

        New state:  Down

           Action:  No actions are necessary.  For example, the
                    interface variables have already been reset upon
                    entering the Loopback state.  Note that reception of



Moy                         Standards Track                    [Page 74]

RFC 2328                     OSPF Version 2                   April 1998


                    an InterfaceUp event is necessary before the
                    interface again becomes fully functional.


    9.4.  Electing the Designated Router



        This section describes the algorithm used for calculating a
        network's Designated Router and Backup Designated Router.  This
        algorithm is invoked by the Interface state machine.  The
        initial time a router runs the election algorithm for a network,
        the network's Designated Router and Backup Designated Router are
        initialized to 0.0.0.0.  This indicates the lack of both a
        Designated Router and a Backup Designated Router.

        The Designated Router election algorithm proceeds as follows:
        Call the router doing the calculation Router X.  The list of
        neighbors attached to the network and having established
        bidirectional communication with Router X is examined.  This
        list is precisely the collection of Router X's neighbors (on
        this network) whose state is greater than or equal to 2-Way (see
        Section 10.1).  Router X itself is also considered to be on the
        list.  Discard all routers from the list that are ineligible to
        become Designated Router.  (Routers having Router Priority of 0
        are ineligible to become Designated Router.)  The following
        steps are then executed, considering only those routers that
        remain on the list:

        (1) Note the current values for the network's Designated Router
            and Backup Designated Router.  This is used later for
            comparison purposes.

        (2) Calculate the new Backup Designated Router for the network
            as follows.  Only those routers on the list that have not
            declared themselves to be Designated Router are eligible to
            become Backup Designated Router.  If one or more of these
            routers have declared themselves Backup Designated Router
            (i.e., they are currently listing themselves as Backup
            Designated Router, but not as Designated Router, in their
            Hello Packets) the one having highest Router Priority is
            declared to be Backup Designated Router.  In case of a tie,
            the one having the highest Router ID is chosen.  If no
            routers have declared themselves Backup Designated Router,



Moy                         Standards Track                    [Page 75]

RFC 2328                     OSPF Version 2                   April 1998


            choose the router having highest Router Priority, (again
            excluding those routers who have declared themselves
            Designated Router), and again use the Router ID to break
            ties.

        (3) Calculate the new Designated Router for the network as
            follows.  If one or more of the routers have declared
            themselves Designated Router (i.e., they are currently
            listing themselves as Designated Router in their Hello
            Packets) the one having highest Router Priority is declared
            to be Designated Router.  In case of a tie, the one having
            the highest Router ID is chosen.  If no routers have
            declared themselves Designated Router, assign the Designated
            Router to be the same as the newly elected Backup Designated
            Router.

        (4) If Router X is now newly the Designated Router or newly the
            Backup Designated Router, or is now no longer the Designated
            Router or no longer the Backup Designated Router, repeat
            steps 2 and 3, and then proceed to step 5.  For example, if
            Router X is now the Designated Router, when step 2 is
            repeated X will no longer be eligible for Backup Designated
            Router election.  Among other things, this will ensure that
            no router will declare itself both Backup Designated Router
            and Designated Router.[5]

        (5) As a result of these calculations, the router itself may now
            be Designated Router or Backup Designated Router.  See
            Sections 7.3 and 7.4 for the additional duties this would
            entail.  The router's interface state should be set
            accordingly.  If the router itself is now Designated Router,
            the new interface state is DR.  If the router itself is now
            Backup Designated Router, the new interface state is Backup.
            Otherwise, the new interface state is DR Other.

        (6) If the attached network is an NBMA network, and the router
            itself has just become either Designated Router or Backup
            Designated Router, it must start sending Hello Packets to
            those neighbors that are not eligible to become Designated
            Router (see Section 9.5.1).  This is done by invoking the
            neighbor event Start for each neighbor having a Router
            Priority of 0.



Moy                         Standards Track                    [Page 76]

RFC 2328                     OSPF Version 2                   April 1998


        (7) If the above calculations have caused the identity of either
            the Designated Router or Backup Designated Router to change,
            the set of adjacencies associated with this interface will
            need to be modified.  Some adjacencies may need to be
            formed, and others may need to be broken.  To accomplish
            this, invoke the event AdjOK?  on all neighbors whose state
            is at least 2-Way.  This will cause their eligibility for
            adjacency to be reexamined (see Sections 10.3 and 10.4).


        The reason behind the election algorithm's complexity is the
        desire for an orderly transition from Backup Designated Router
        to Designated Router, when the current Designated Router fails.
        This orderly transition is ensured through the introduction of
        hysteresis: no new Backup Designated Router can be chosen until
        the old Backup accepts its new Designated Router
        responsibilities.

        The above procedure may elect the same router to be both
        Designated Router and Backup Designated Router, although that
        router will never be the calculating router (Router X) itself.
        The elected Designated Router may not be the router having the
        highest Router Priority, nor will the Backup Designated Router
        necessarily have the second highest Router Priority.  If Router
        X is not itself eligible to become Designated Router, it is
        possible that neither a Backup Designated Router nor a
        Designated Router will be selected in the above procedure.  Note
        also that if Router X is the only attached router that is
        eligible to become Designated Router, it will select itself as
        Designated Router and there will be no Backup Designated Router
        for the network.


    9.5.  Sending Hello packets



        Hello packets are sent out each functioning router interface.
        They are used to discover and maintain neighbor
        relationships.[6] On broadcast and NBMA networks, Hello Packets
        are also used to elect the Designated Router and Backup
        Designated Router.





Moy                         Standards Track                    [Page 77]

RFC 2328                     OSPF Version 2                   April 1998


        The format of an Hello packet is detailed in Section A.3.2.  The
        Hello Packet contains the router's Router Priority (used in
        choosing the Designated Router), and the interval between Hello
        Packets sent out the interface (HelloInterval).  The Hello
        Packet also indicates how often a neighbor must be heard from to
        remain active (RouterDeadInterval).  Both HelloInterval and
        RouterDeadInterval must be the same for all routers attached to
        a common network.  The Hello packet also contains the IP address
        mask of the attached network (Network Mask).  On unnumbered
        point-to-point networks and on virtual links this field should
        be set to 0.0.0.0.

        The Hello packet's Options field describes the router's optional
        OSPF capabilities.  One optional capability is defined in this
        specification (see Sections 4.5 and A.2).  The E-bit of the
        Options field should be set if and only if the attached area is
        capable of processing AS-external-LSAs (i.e., it is not a stub
        area).  If the E-bit is set incorrectly the neighboring routers
        will refuse to accept the Hello Packet (see Section 10.5).
        Unrecognized bits in the Hello Packet's Options field should be
        set to zero.

        In order to ensure two-way communication between adjacent
        routers, the Hello packet contains the list of all routers on
        the network from which Hello Packets have been seen recently.
        The Hello packet also contains the router's current choice for
        Designated Router and Backup Designated Router.  A value of
        0.0.0.0 in these fields means that one has not yet been
        selected.

        On broadcast networks and physical point-to-point networks,
        Hello packets are sent every HelloInterval seconds to the IP
        multicast address AllSPFRouters.  On virtual links, Hello
        packets are sent as unicasts (addressed directly to the other
        end of the virtual link) every HelloInterval seconds. On Point-
        to-MultiPoint networks, separate Hello packets are sent to each
        attached neighbor every HelloInterval seconds. Sending of Hello
        packets on NBMA networks is covered in the next section.







Moy                         Standards Track                    [Page 78]

RFC 2328                     OSPF Version 2                   April 1998


        9.5.1.  Sending Hello packets on NBMA networks



            Static configuration information may be necessary in order
            for the Hello Protocol to function on non-broadcast networks
            (see Sections C.5 and C.6).  On NBMA networks, every
            attached router which is eligible to become Designated
            Router becomes aware of all of its neighbors on the network
            (either through configuration or by some unspecified
            mechanism).  Each neighbor is labelled with the neighbor's
            Designated Router eligibility.

            The interface state must be at least Waiting for any Hello
            Packets to be sent out the NBMA interface.  Hello Packets
            are then sent directly (as unicasts) to some subset of a
            router's neighbors.  Sometimes an Hello Packet is sent
            periodically on a timer; at other times it is sent as a
            response to a received Hello Packet.  A router's hello-
            sending behavior varies depending on whether the router
            itself is eligible to become Designated Router.

            If the router is eligible to become Designated Router, it
            must periodically send Hello Packets to all neighbors that
            are also eligible.  In addition, if the router is itself the
            Designated Router or Backup Designated Router, it must also
            send periodic Hello Packets to all other neighbors.  This
            means that any two eligible routers are always exchanging
            Hello Packets, which is necessary for the correct operation
            of the Designated Router election algorithm.  To minimize
            the number of Hello Packets sent, the number of eligible
            routers on an NBMA network should be kept small.

            If the router is not eligible to become Designated Router,
            it must periodically send Hello Packets to both the
            Designated Router and the Backup Designated Router (if they
            exist).  It must also send an Hello Packet in reply to an
            Hello Packet received from any eligible neighbor (other than
            the current Designated Router and Backup Designated Router).
            This is needed to establish an initial bidirectional
            relationship with any potential Designated Router.

            When sending Hello packets periodically to any neighbor, the
            interval between Hello Packets is determined by the



Moy                         Standards Track                    [Page 79]

RFC 2328                     OSPF Version 2                   April 1998


            neighbor's state.  If the neighbor is in state Down, Hello
            Packets are sent every PollInterval seconds.  Otherwise,
            Hello Packets are sent every HelloInterval seconds.


10.  The Neighbor Data Structure



    An OSPF router converses with its neighboring routers.  Each
    separate conversation is described by a "neighbor data structure".
    Each conversation is bound to a particular OSPF router interface,
    and is identified either by the neighboring router's OSPF Router ID
    or by its Neighbor IP address (see below).  Thus if the OSPF router
    and another router have multiple attached networks in common,
    multiple conversations ensue, each described by a unique neighbor
    data structure.  Each separate conversation is loosely referred to
    in the text as being a separate "neighbor".

    The neighbor data structure contains all information pertinent to
    the forming or formed adjacency between the two neighbors.
    (However, remember that not all neighbors become adjacent.)  An
    adjacency can be viewed as a highly developed conversation between
    two routers.


    State
        The functional level of the neighbor conversation.  This is
        described in more detail in Section 10.1.

    Inactivity Timer
        A single shot timer whose firing indicates that no Hello Packet
        has been seen from this neighbor recently.  The length of the
        timer is RouterDeadInterval seconds.

    Master/Slave
        When the two neighbors are exchanging databases, they form a
        master/slave relationship.  The master sends the first Database
        Description Packet, and is the only part that is allowed to
        retransmit.  The slave can only respond to the master's Database
        Description Packets.  The master/slave relationship is
        negotiated in state ExStart.





Moy                         Standards Track                    [Page 80]

RFC 2328                     OSPF Version 2                   April 1998


    DD Sequence Number
        The DD Sequence number of the Database Description packet that
        is currently being sent to the neighbor.

    Last received Database Description packet
        The initialize(I), more (M) and master(MS) bits, Options field,
        and DD sequence number contained in the last Database
        Description packet received from the neighbor. Used to determine
        whether the next Database Description packet received from the
        neighbor is a duplicate.

    Neighbor ID
        The OSPF Router ID of the neighboring router.  The Neighbor ID
        is learned when Hello packets are received from the neighbor, or
        is configured if this is a virtual adjacency (see Section C.4).

    Neighbor Priority
        The Router Priority of the neighboring router.  Contained in the
        neighbor's Hello packets, this item is used when selecting the
        Designated Router for the attached network.

    Neighbor IP address
        The IP address of the neighboring router's interface to the
        attached network.  Used as the Destination IP address when
        protocol packets are sent as unicasts along this adjacency.
        Also used in router-LSAs as the Link ID for the attached network
        if the neighboring router is selected to be Designated Router
        (see Section 12.4.1).  The Neighbor IP address is learned when
        Hello packets are received from the neighbor.  For virtual
        links, the Neighbor IP address is learned during the routing
        table build process (see Section 15).

    Neighbor Options
        The optional OSPF capabilities supported by the neighbor.
        Learned during the Database Exchange process (see Section 10.6).
        The neighbor's optional OSPF capabilities are also listed in its
        Hello packets.  This enables received Hello Packets to be
        rejected (i.e., neighbor relationships will not even start to
        form) if there is a mismatch in certain crucial OSPF
        capabilities (see Section 10.5).  The optional OSPF capabilities
        are documented in Section 4.5.




Moy                         Standards Track                    [Page 81]

RFC 2328                     OSPF Version 2                   April 1998


    Neighbor's Designated Router
        The neighbor's idea of the Designated Router.  If this is the
        neighbor itself, this is important in the local calculation of
        the Designated Router.  Defined only on broadcast and NBMA
        networks.

    Neighbor's Backup Designated Router
        The neighbor's idea of the Backup Designated Router.  If this is
        the neighbor itself, this is important in the local calculation
        of the Backup Designated Router.  Defined only on broadcast and
        NBMA networks.


    The next set of variables are lists of LSAs.  These lists describe
    subsets of the area link-state database.  This memo defines five
    distinct types of LSAs, all of which may be present in an area
    link-state database: router-LSAs, network-LSAs, and Type 3 and 4
    summary-LSAs (all stored in the area data structure), and AS-
    external-LSAs (stored in the global data structure).


    Link state retransmission list
        The list of LSAs that have been flooded but not acknowledged on
        this adjacency.  These will be retransmitted at intervals until
        they are acknowledged, or until the adjacency is destroyed.

    Database summary list
        The complete list of LSAs that make up the area link-state
        database, at the moment the neighbor goes into Database Exchange
        state.  This list is sent to the neighbor in Database
        Description packets.

    Link state request list
        The list of LSAs that need to be received from this neighbor in
        order to synchronize the two neighbors' link-state databases.
        This list is created as Database Description packets are
        received, and is then sent to the neighbor in Link State Request
        packets.  The list is depleted as appropriate Link State Update
        packets are received.






Moy                         Standards Track                    [Page 82]

RFC 2328                     OSPF Version 2                   April 1998


    10.1.  Neighbor states



        The state of a neighbor (really, the state of a conversation
        being held with a neighboring router) is documented in the
        following sections.  The states are listed in order of
        progressing functionality.  For example, the inoperative state
        is listed first, followed by a list of intermediate states
        before the final, fully functional state is achieved.  The
        specification makes use of this ordering by sometimes making
        references such as "those neighbors/adjacencies in state greater
        than X".  Figures 12 and 13 show the graph of neighbor state
        changes.  The arcs of the graphs are labelled with the event
        causing the state change.  The neighbor events are documented in
        Section 10.2.

        The graph in Figure 12 shows the state changes effected by the
        Hello Protocol.  The Hello Protocol is responsible for neighbor
        acquisition and maintenance, and for ensuring two way
        communication between neighbors.

        The graph in Figure 13 shows the forming of an adjacency.  Not
        every two neighboring routers become adjacent (see Section
        10.4).  The adjacency starts to form when the neighbor is in
        state ExStart.  After the two routers discover their
        master/slave status, the state transitions to Exchange.  At this
        point the neighbor starts to be used in the flooding procedure,
        and the two neighboring routers begin synchronizing their
        databases.  When this synchronization is finished, the neighbor
        is in state Full and we say that the two routers are fully
        adjacent.  At this point the adjacency is listed in LSAs.

        For a more detailed description of neighbor state changes,
        together with the additional actions involved in each change,
        see Section 10.3.


        Down
            This is the initial state of a neighbor conversation.  It
            indicates that there has been no recent information received
            from the neighbor.  On NBMA networks, Hello packets may
            still be sent to "Down" neighbors, although at a reduced
            frequency (see Section 9.5.1).



Moy                         Standards Track                    [Page 83]

RFC 2328                     OSPF Version 2                   April 1998



                                   +----+
                                   |Down|
                                   +----+
                                     |\
                                     | \Start
                                     |  \      +-------+
                             Hello   |   +---->|Attempt|
                            Received |         +-------+
                                     |             |
                             +----+<-+             |HelloReceived
                             |Init|<---------------+
                             +----+<--------+
                                |           |
                                |2-Way      |1-Way
                                |Received   |Received
                                |           |
              +-------+         |        +-----+
              |ExStart|<--------+------->|2-Way|
              +-------+                  +-----+

              Figure 12: Neighbor state changes (Hello Protocol)

                  In addition to the state transitions pictured,
                  Event KillNbr always forces Down State,
                  Event InactivityTimer always forces Down State,
                  Event LLDown always forces Down State


















Moy                         Standards Track                    [Page 84]

RFC 2328                     OSPF Version 2                   April 1998


                                  +-------+
                                  |ExStart|
                                  +-------+
                                    |
                     NegotiationDone|
                                    +->+--------+
                                       |Exchange|
                                    +--+--------+
                                    |
                            Exchange|
                              Done  |
                    +----+          |      +-------+
                    |Full|<---------+----->|Loading|
                    +----+<-+              +-------+
                            |  LoadingDone     |
                            +------------------+

            Figure 13: Neighbor state changes (Database Exchange)

                In addition to the state transitions pictured,
                Event SeqNumberMismatch forces ExStart state,
                Event BadLSReq forces ExStart state,
                Event 1-Way forces Init state,
                Event KillNbr always forces Down State,
                Event InactivityTimer always forces Down State,
                Event LLDown always forces Down State,
                Event AdjOK? leads to adjacency forming/breaking

        Attempt
            This state is only valid for neighbors attached to NBMA
            networks.  It indicates that no recent information has been
            received from the neighbor, but that a more concerted effort
            should be made to contact the neighbor.  This is done by
            sending the neighbor Hello packets at intervals of
            HelloInterval (see Section 9.5.1).

        Init
            In this state, an Hello packet has recently been seen from
            the neighbor.  However, bidirectional communication has not
            yet been established with the neighbor (i.e., the router
            itself did not appear in the neighbor's Hello packet).  All




Moy                         Standards Track                    [Page 85]

RFC 2328                     OSPF Version 2                   April 1998


            neighbors in this state (or higher) are listed in the Hello
            packets sent from the associated interface.

        2-Way
            In this state, communication between the two routers is
            bidirectional.  This has been assured by the operation of
            the Hello Protocol.  This is the most advanced state short
            of beginning adjacency establishment.  The (Backup)
            Designated Router is selected from the set of neighbors in
            state 2-Way or greater.

        ExStart
            This is the first step in creating an adjacency between the
            two neighboring routers.  The goal of this step is to decide
            which router is the master, and to decide upon the initial
            DD sequence number.  Neighbor conversations in this state or
            greater are called adjacencies.

        Exchange
            In this state the router is describing its entire link state
            database by sending Database Description packets to the
            neighbor.  Each Database Description Packet has a DD
            sequence number, and is explicitly acknowledged.  Only one
            Database Description Packet is allowed outstanding at any
            one time.  In this state, Link State Request Packets may
            also be sent asking for the neighbor's more recent LSAs.
            All adjacencies in Exchange state or greater are used by the
            flooding procedure.  In fact, these adjacencies are fully
            capable of transmitting and receiving all types of OSPF
            routing protocol packets.

        Loading
            In this state, Link State Request packets are sent to the
            neighbor asking for the more recent LSAs that have been
            discovered (but not yet received) in the Exchange state.

        Full
            In this state, the neighboring routers are fully adjacent.
            These adjacencies will now appear in router-LSAs and
            network-LSAs.





Moy                         Standards Track                    [Page 86]

RFC 2328                     OSPF Version 2                   April 1998


    10.2.  Events causing neighbor state changes



        State changes can be effected by a number of events.  These
        events are shown in the labels of the arcs in Figures 12 and 13.
        The label definitions are as follows:


        HelloReceived
            An Hello packet has been received from the neighbor.

        Start
            This is an indication that Hello Packets should now be sent
            to the neighbor at intervals of HelloInterval seconds.  This
            event is generated only for neighbors associated with NBMA
            networks.

        2-WayReceived
            Bidirectional communication has been realized between the
            two neighboring routers.  This is indicated by the router
            seeing itself in the neighbor's Hello packet.

        NegotiationDone
            The Master/Slave relationship has been negotiated, and DD
            sequence numbers have been exchanged.  This signals the
            start of the sending/receiving of Database Description
            packets.  For more information on the generation of this
            event, consult Section 10.8.

        ExchangeDone
            Both routers have successfully transmitted a full sequence
            of Database Description packets.  Each router now knows what
            parts of its link state database are out of date.  For more
            information on the generation of this event, consult Section
            10.8.

        BadLSReq
            A Link State Request has been received for an LSA not
            contained in the database.  This indicates an error in the
            Database Exchange process.

        Loading Done
            Link State Updates have been received for all out-of-date



Moy                         Standards Track                    [Page 87]

RFC 2328                     OSPF Version 2                   April 1998


            portions of the database.  This is indicated by the Link
            state request list becoming empty after the Database
            Exchange process has completed.

        AdjOK?
            A decision must be made as to whether an adjacency should be
            established/maintained with the neighbor.  This event will
            start some adjacencies forming, and destroy others.


        The following events cause well developed neighbors to revert to
        lesser states.  Unlike the above events, these events may occur
        when the neighbor conversation is in any of a number of states.


        SeqNumberMismatch
            A Database Description packet has been received that either
            a) has an unexpected DD sequence number, b) unexpectedly has
            the Init bit set or c) has an Options field differing from
            the last Options field received in a Database Description
            packet.  Any of these conditions indicate that some error
            has occurred during adjacency establishment.

        1-Way
            An Hello packet has been received from the neighbor, in
            which the router is not mentioned.  This indicates that
            communication with the neighbor is not bidirectional.

        KillNbr
            This  is  an  indication that  all  communication  with  the
            neighbor  is now  impossible,  forcing  the  neighbor  to
            revert  to  Down  state.

        InactivityTimer
            The inactivity Timer has fired.  This means that no Hello
            packets have been seen recently from the neighbor.  The
            neighbor reverts to Down state.

        LLDown
            This is an indication from the lower level protocols that
            the neighbor is now unreachable.  For example, on an X.25
            network this could be indicated by an X.25 clear indication



Moy                         Standards Track                    [Page 88]

RFC 2328                     OSPF Version 2                   April 1998


            with appropriate cause and diagnostic fields.  This event
            forces the neighbor into Down state.


    10.3.  The Neighbor state machine



        A detailed description of the neighbor state changes follows.
        Each state change is invoked by an event (Section 10.2).  This
        event may produce different effects, depending on the current
        state of the neighbor.  For this reason, the state machine below
        is organized by current neighbor state and received event.  Each
        entry in the state machine describes the resulting new neighbor
        state and the required set of additional actions.

        When a neighbor's state changes, it may be necessary to rerun
        the Designated Router election algorithm.  This is determined by
        whether the interface NeighborChange event is generated (see
        Section 9.2).  Also, if the Interface is in DR state (the router
        is itself Designated Router), changes in neighbor state may
        cause a new network-LSA to be originated (see Section 12.4).

        When the neighbor state machine needs to invoke the interface
        state machine, it should be done as a scheduled task (see
        Section 4.4).  This simplifies things, by ensuring that neither
        state machine will be executed recursively.


         State(s):  Down

            Event:  Start

        New state:  Attempt

           Action:  Send an Hello Packet to the neighbor (this neighbor
                    is always associated with an NBMA network) and start
                    the Inactivity Timer for the neighbor.  The timer's
                    later firing would indicate that communication with
                    the neighbor was not attained.


         State(s):  Attempt




Moy                         Standards Track                    [Page 89]

RFC 2328                     OSPF Version 2                   April 1998


            Event:  HelloReceived

        New state:  Init

           Action:  Restart the Inactivity Timer for the neighbor, since
                    the neighbor has now been heard from.


         State(s):  Down

            Event:  HelloReceived

        New state:  Init

           Action:  Start the Inactivity Timer for the neighbor.  The
                    timer's later firing would indicate that the
                    neighbor is dead.


         State(s):  Init or greater

            Event:  HelloReceived

        New state:  No state change.

           Action:  Restart the Inactivity Timer for the neighbor, since
                    the neighbor has again been heard from.


         State(s):  Init

            Event:  2-WayReceived

        New state:  Depends upon action routine.

           Action:  Determine whether an adjacency should be established
                    with the neighbor (see Section 10.4).  If not, the
                    new neighbor state is 2-Way.

                    Otherwise (an adjacency should be established) the
                    neighbor state transitions to ExStart.  Upon
                    entering this state, the router increments the DD



Moy                         Standards Track                    [Page 90]

RFC 2328                     OSPF Version 2                   April 1998


                    sequence number in the neighbor data structure.  If
                    this is the first time that an adjacency has been
                    attempted, the DD sequence number should be assigned
                    some unique value (like the time of day clock).  It
                    then declares itself master (sets the master/slave
                    bit to master), and starts sending Database
                    Description Packets, with the initialize (I), more
                    (M) and master (MS) bits set.  This Database
                    Description Packet should be otherwise empty.  This
                    Database Description Packet should be retransmitted
                    at intervals of RxmtInterval until the next state is
                    entered (see Section 10.8).


         State(s):  ExStart

            Event:  NegotiationDone

        New state:  Exchange

           Action:  The router must list the contents of its entire area
                    link state database in the neighbor Database summary
                    list.  The area link state database consists of the
                    router-LSAs, network-LSAs and summary-LSAs contained
                    in the area structure, along with the AS-external-
                    LSAs contained in the global structure.  AS-
                    external-LSAs are omitted from a virtual neighbor's
                    Database summary list.  AS-external-LSAs are omitted
                    from the Database summary list if the area has been
                    configured as a stub (see Section 3.6).  LSAs whose
                    age is equal to MaxAge are instead added to the
                    neighbor's Link state retransmission list.  A
                    summary of the Database summary list will be sent to
                    the neighbor in Database Description packets.  Each
                    Database Description Packet has a DD sequence
                    number, and is explicitly acknowledged.  Only one
                    Database Description Packet is allowed outstanding
                    at any one time.  For more detail on the sending and
                    receiving of Database Description packets, see
                    Sections 10.8 and 10.6.





Moy                         Standards Track                    [Page 91]

RFC 2328                     OSPF Version 2                   April 1998


         State(s):  Exchange

            Event:  ExchangeDone

        New state:  Depends upon action routine.

           Action:  If the neighbor Link state request list is empty,
                    the new neighbor state is Full.  No other action is
                    required.  This is an adjacency's final state.

                    Otherwise, the new neighbor state is Loading.  Start
                    (or continue) sending Link State Request packets to
                    the neighbor (see Section 10.9).  These are requests
                    for the neighbor's more recent LSAs (which were
                    discovered but not yet received in the Exchange
                    state).  These LSAs are listed in the Link state
                    request list associated with the neighbor.


         State(s):  Loading

            Event:  Loading Done

        New state:  Full

           Action:  No action required.  This is an adjacency's final
                    state.


         State(s):  2-Way

            Event:  AdjOK?

        New state:  Depends upon action routine.

           Action:  Determine whether an adjacency should be formed with
                    the neighboring router (see Section 10.4).  If not,
                    the neighbor state remains at 2-Way.  Otherwise,
                    transition the neighbor state to ExStart and perform
                    the actions associated with the above state machine
                    entry for state Init and event 2-WayReceived.




Moy                         Standards Track                    [Page 92]

RFC 2328                     OSPF Version 2                   April 1998


         State(s):  ExStart or greater

            Event:  AdjOK?

        New state:  Depends upon action routine.

           Action:  Determine whether the neighboring router should
                    still be adjacent.  If yes, there is no state change
                    and no further action is necessary.

                    Otherwise, the (possibly partially formed) adjacency
                    must be destroyed.  The neighbor state transitions
                    to 2-Way.  The Link state retransmission list,
                    Database summary list and Link state request list
                    are cleared of LSAs.


         State(s):  Exchange or greater

            Event:  SeqNumberMismatch

        New state:  ExStart

           Action:  The (possibly partially formed) adjacency is torn
                    down, and then an attempt is made at
                    reestablishment.  The neighbor state first
                    transitions to ExStart.  The Link state
                    retransmission list, Database summary list and Link
                    state request list are cleared of LSAs.  Then the
                    router increments the DD sequence number in the
                    neighbor data structure, declares itself master
                    (sets the master/slave bit to master), and starts
                    sending Database Description Packets, with the
                    initialize (I), more (M) and master (MS) bits set.
                    This Database Description Packet should be otherwise
                    empty (see Section 10.8).


         State(s):  Exchange or greater

            Event:  BadLSReq




Moy                         Standards Track                    [Page 93]

RFC 2328                     OSPF Version 2                   April 1998


        New state:  ExStart

           Action:  The action for event BadLSReq is exactly the same as
                    for the neighbor event SeqNumberMismatch.  The
                    (possibly partially formed) adjacency is torn down,
                    and then an attempt is made at reestablishment.  For
                    more information, see the neighbor state machine
                    entry that is invoked when event SeqNumberMismatch
                    is generated in state Exchange or greater.


         State(s):  Any state

            Event:  KillNbr

        New state:  Down

           Action:  The Link state retransmission list, Database summary
                    list and Link state request list are cleared of
                    LSAs.  Also, the Inactivity Timer is disabled.


         State(s):  Any state

            Event:  LLDown

        New state:  Down

           Action:  The Link state retransmission list, Database summary
                    list and Link state request list are cleared of
                    LSAs.  Also, the Inactivity Timer is disabled.


         State(s):  Any state

            Event:  InactivityTimer

        New state:  Down

           Action:  The Link state retransmission list, Database summary
                    list and Link state request list are cleared of
                    LSAs.



Moy                         Standards Track                    [Page 94]

RFC 2328                     OSPF Version 2                   April 1998


         State(s):  2-Way or greater

            Event:  1-WayReceived

        New state:  Init

           Action:  The Link state retransmission list, Database summary
                    list and Link state request list are cleared of
                    LSAs.


         State(s):  2-Way or greater

            Event:  2-WayReceived

        New state:  No state change.

           Action:  No action required.


         State(s):  Init

            Event:  1-WayReceived

        New state:  No state change.

           Action:  No action required.


    10.4.  Whether to become adjacent



        Adjacencies are established with some subset of the router's
        neighbors.  Routers connected by point-to-point networks,
        Point-to-MultiPoint networks and virtual links always become
        adjacent.  On broadcast and NBMA networks, all routers become
        adjacent to both the Designated Router and the Backup Designated
        Router.

        The adjacency-forming decision occurs in two places in the
        neighbor state machine.  First, when bidirectional communication
        is initially established with the neighbor, and secondly, when
        the identity of the attached network's (Backup) Designated



Moy                         Standards Track                    [Page 95]

RFC 2328                     OSPF Version 2                   April 1998


        Router changes.  If the decision is made to not attempt an
        adjacency, the state of the neighbor communication stops at 2-
        Way.

        An adjacency should be established with a bidirectional neighbor
        when at least one of the following conditions holds:


        o   The underlying network type is point-to-point

        o   The underlying network type is Point-to-MultiPoint

        o   The underlying network type is virtual link

        o   The router itself is the Designated Router

        o   The router itself is the Backup Designated Router

        o   The neighboring router is the Designated Router

        o   The neighboring router is the Backup Designated Router


    10.5.  Receiving Hello Packets



        This section explains the detailed processing of a received
        Hello Packet.  (See Section A.3.2 for the format of Hello
        packets.)  The generic input processing of OSPF packets will
        have checked the validity of the IP header and the OSPF packet
        header.  Next, the values of the Network Mask, HelloInterval,
        and RouterDeadInterval fields in the received Hello packet must
        be checked against the values configured for the receiving
        interface.  Any mismatch causes processing to stop and the
        packet to be dropped.  In other words, the above fields are
        really describing the attached network's configuration. However,
        there is one exception to the above rule: on point-to-point
        networks and on virtual links, the Network Mask in the received
        Hello Packet should be ignored.

        The receiving interface attaches to a single OSPF area (this
        could be the backbone).  The setting of the E-bit found in the
        Hello Packet's Options field must match this area's



Moy                         Standards Track                    [Page 96]

RFC 2328                     OSPF Version 2                   April 1998


        ExternalRoutingCapability.  If AS-external-LSAs are not flooded
        into/throughout the area (i.e, the area is a "stub") the E-bit
        must be clear in received Hello Packets, otherwise the E-bit
        must be set.  A mismatch causes processing to stop and the
        packet to be dropped.  The setting of the rest of the bits in
        the Hello Packet's Options field should be ignored.

        At this point, an attempt is made to match the source of the
        Hello Packet to one of the receiving interface's neighbors.  If
        the receiving interface connects to a broadcast, Point-to-
        MultiPoint or NBMA network the source is identified by the IP
        source address found in the Hello's IP header.  If the receiving
        interface connects to a point-to-point link or a virtual link,
        the source is identified by the Router ID found in the Hello's
        OSPF packet header.  The interface's current list of neighbors
        is contained in the interface's data structure.  If a matching
        neighbor structure cannot be found, (i.e., this is the first
        time the neighbor has been detected), one is created.  The
        initial state of a newly created neighbor is set to Down.

        When receiving an Hello Packet from a neighbor on a broadcast,
        Point-to-MultiPoint or NBMA network, set the neighbor
        structure's Neighbor ID equal to the Router ID found in the
        packet's OSPF header.  For these network types, the neighbor
        structure's Router Priority field, Neighbor's Designated Router
        field, and Neighbor's Backup Designated Router field are also
        set equal to the corresponding fields found in the received
        Hello Packet; changes in these fields should be noted for
        possible use in the steps below.  When receiving an Hello on a
        point-to-point network (but not on a virtual link) set the
        neighbor structure's Neighbor IP address to the packet's IP
        source address.

        Now the rest of the Hello Packet is examined, generating events
        to be given to the neighbor and interface state machines.  These
        state machines are specified either to be executed or scheduled
        (see Section 4.4).  For example, by specifying below that the
        neighbor state machine be executed in line, several neighbor
        state transitions may be effected by a single received Hello:






Moy                         Standards Track                    [Page 97]

RFC 2328                     OSPF Version 2                   April 1998


        o   Each Hello Packet causes the neighbor state machine to be
            executed with the event HelloReceived.

        o   Then the list of neighbors contained in the Hello Packet is
            examined.  If the router itself appears in this list, the
            neighbor state machine should be executed with the event 2-
            WayReceived.  Otherwise, the neighbor state machine should
            be executed with the event 1-WayReceived, and the processing
            of the packet stops.

        o   Next, if a change in the neighbor's Router Priority field
            was noted, the receiving interface's state machine is
            scheduled with the event NeighborChange.

        o   If the neighbor is both declaring itself to be Designated
            Router (Hello Packet's Designated Router field = Neighbor IP
            address) and the Backup Designated Router field in the
            packet is equal to 0.0.0.0 and the receiving interface is in
            state Waiting, the receiving interface's state machine is
            scheduled with the event BackupSeen.  Otherwise, if the
            neighbor is declaring itself to be Designated Router and it
            had not previously, or the neighbor is not declaring itself
            Designated Router where it had previously, the receiving
            interface's state machine is scheduled with the event
            NeighborChange.

        o   If the neighbor is declaring itself to be Backup Designated
            Router (Hello Packet's Backup Designated Router field =
            Neighbor IP address) and the receiving interface is in state
            Waiting, the receiving interface's state machine is
            scheduled with the event BackupSeen.  Otherwise, if the
            neighbor is declaring itself to be Backup Designated Router
            and it had not previously, or the neighbor is not declaring
            itself Backup Designated Router where it had previously, the
            receiving interface's state machine is scheduled with the
            event NeighborChange.

        On NBMA networks, receipt of an Hello Packet may also cause an
        Hello Packet to be sent back to the neighbor in response. See
        Section 9.5.1 for more details.





Moy                         Standards Track                    [Page 98]

RFC 2328                     OSPF Version 2                   April 1998


    10.6.  Receiving Database Description Packets



        This section explains the detailed processing of a received
        Database Description Packet.  The incoming Database Description
        Packet has already been associated with a neighbor and receiving
        interface by the generic input packet processing (Section 8.2).
        Whether the Database Description packet should be accepted, and
        if so, how it should be further processed depends upon the
        neighbor state.

        If a Database Description packet is accepted, the following
        packet fields should be saved in the corresponding neighbor data
        structure under "last received Database Description packet":
        the packet's initialize(I), more (M) and master(MS) bits,
        Options field, and DD sequence number. If these fields are set
        identically in two consecutive Database Description packets
        received from the neighbor, the second Database Description
        packet is considered to be a "duplicate" in the processing
        described below.

        If the Interface MTU field in the Database Description packet
        indicates an IP datagram size that is larger than the router can
        accept on the receiving interface without fragmentation, the
        Database Description packet is rejected.  Otherwise, if the
        neighbor state is:

        Down
            The packet should be rejected.

        Attempt
            The packet should be rejected.

        Init
            The neighbor state machine should be executed with the event
            2-WayReceived.  This causes an immediate state change to
            either state 2-Way or state ExStart. If the new state is
            ExStart, the processing of the current packet should then
            continue in this new state by falling through to case
            ExStart below.






Moy                         Standards Track                    [Page 99]

RFC 2328                     OSPF Version 2                   April 1998


        2-Way
            The packet should be ignored.  Database Description Packets
            are used only for the purpose of bringing up adjacencies.[7]

        ExStart
            If the received packet matches one of the following cases,
            then the neighbor state machine should be executed with the
            event NegotiationDone (causing the state to transition to
            Exchange), the packet's Options field should be recorded in
            the neighbor structure's Neighbor Options field and the
            packet should be accepted as next in sequence and processed
            further (see below).  Otherwise, the packet should be
            ignored.

            o   The initialize(I), more (M) and master(MS) bits are set,
                the contents of the packet are empty, and the neighbor's
                Router ID is larger than the router's own.  In this case
                the router is now Slave.  Set the master/slave bit to
                slave, and set the neighbor data structure's DD sequence
                number to that specified by the master.

            o   The initialize(I) and master(MS) bits are off, the
                packet's DD sequence number equals the neighbor data
                structure's DD sequence number (indicating
                acknowledgment) and the neighbor's Router ID is smaller
                than the router's own.  In this case the router is
                Master.

        Exchange
            Duplicate Database Description packets are discarded by the
            master, and cause the slave to retransmit the last Database
            Description packet that it had sent. Otherwise (the packet
            is not a duplicate):

            o   If the state of the MS-bit is inconsistent with the
                master/slave state of the connection, generate the
                neighbor event SeqNumberMismatch and stop processing the
                packet.

            o   If the initialize(I) bit is set, generate the neighbor
                event SeqNumberMismatch and stop processing the packet.




Moy                         Standards Track                   [Page 100]

RFC 2328                     OSPF Version 2                   April 1998


            o   If the packet's Options field indicates a different set
                of optional OSPF capabilities than were previously
                received from the neighbor (recorded in the Neighbor
                Options field of the neighbor structure), generate the
                neighbor event SeqNumberMismatch and stop processing the
                packet.

            o   Database Description packets must be processed in
                sequence, as indicated by the packets' DD sequence
                numbers. If the router is master, the next packet
                received should have DD sequence number equal to the DD
                sequence number in the neighbor data structure. If the
                router is slave, the next packet received should have DD
                sequence number equal to one more than the DD sequence
                number stored in the neighbor data structure. In either
                case, if the packet is the next in sequence it should be
                accepted and its contents processed as specified below.

            o   Else, generate the neighbor event SeqNumberMismatch and
                stop processing the packet.

        Loading or Full
            In this state, the router has sent and received an entire
            sequence of Database Description Packets.  The only packets
            received should be duplicates (see above).  In particular,
            the packet's Options field should match the set of optional
            OSPF capabilities previously indicated by the neighbor
            (stored in the neighbor structure's Neighbor Options field).
            Any other packets received, including the reception of a
            packet with the Initialize(I) bit set, should generate the
            neighbor event SeqNumberMismatch.[8] Duplicates should be
            discarded by the master.  The slave must respond to
            duplicates by repeating the last Database Description packet
            that it had sent.


        When the router accepts a received Database Description Packet
        as the next in sequence the packet contents are processed as
        follows.  For each LSA listed, the LSA's LS type is checked for
        validity.  If the LS type is unknown (e.g., not one of the LS
        types 1-5 defined by this specification), or if this is an AS-
        external-LSA (LS type = 5) and the neighbor is associated with a



Moy                         Standards Track                   [Page 101]

RFC 2328                     OSPF Version 2                   April 1998


        stub area, generate the neighbor event SeqNumberMismatch and
        stop processing the packet.  Otherwise, the router looks up the
        LSA in its database to see whether it also has an instance of
        the LSA.  If it does not, or if the database copy is less recent
        (see Section 13.1), the LSA is put on the Link state request
        list so that it can be requested (immediately or at some later
        time) in Link State Request Packets.

        When the router accepts a received Database Description Packet
        as the next in sequence, it also performs the following actions,
        depending on whether it is master or slave:


        Master
            Increments the DD sequence number in the neighbor data
            structure.  If the router has already sent its entire
            sequence of Database Description Packets, and the just
            accepted packet has the more bit (M) set to 0, the neighbor
            event ExchangeDone is generated.  Otherwise, it should send
            a new Database Description to the slave.

        Slave
            Sets the DD sequence number in the neighbor data structure
            to the DD sequence number appearing in the received packet.
            The slave must send a Database Description Packet in reply.
            If the received packet has the more bit (M) set to 0, and
            the packet to be sent by the slave will also have the M-bit
            set to 0, the neighbor event ExchangeDone is generated.
            Note that the slave always generates this event before the
            master.


    10.7.  Receiving Link State Request Packets



        This section explains the detailed processing of received Link
        State Request packets.  Received Link State Request Packets
        specify a list of LSAs that the neighbor wishes to receive.
        Link State Request Packets should be accepted when the neighbor
        is in states Exchange, Loading, or Full.  In all other states
        Link State Request Packets should be ignored.





Moy                         Standards Track                   [Page 102]

RFC 2328                     OSPF Version 2                   April 1998


        Each LSA specified in the Link State Request packet should be
        located in the router's database, and copied into Link State
        Update packets for transmission to the neighbor.  These LSAs
        should NOT be placed on the Link state retransmission list for
        the neighbor.  If an LSA cannot be found in the database,
        something has gone wrong with the Database Exchange process, and
        neighbor event BadLSReq should be generated.


    10.8.  Sending Database Description Packets



        This section describes how Database Description Packets are sent
        to a neighbor. The Database Description packet's Interface MTU
        field is set to the size of the largest IP datagram that can be
        sent out the sending interface, without fragmentation.  Common
        MTUs in use in the Internet can be found in Table 7-1 of
        [Ref22]. Interface MTU should be set to 0 in Database
        Description packets sent over virtual links.

        The router's optional OSPF capabilities (see Section 4.5) are
        transmitted to the neighbor in the Options field of the Database
        Description packet.  The router should maintain the same set of
        optional capabilities throughout the Database Exchange and
        flooding procedures.  If for some reason the router's optional
        capabilities change, the Database Exchange procedure should be
        restarted by reverting to neighbor state ExStart.  One optional
        capability is defined in this specification (see Sections 4.5
        and A.2). The E-bit should be set if and only if the attached
        network belongs to a non-stub area. Unrecognized bits in the
        Options field should be set to zero.

        The sending of Database Description packets depends on the
        neighbor's state.  In state ExStart the router sends empty
        Database Description packets, with the initialize (I), more (M)
        and master (MS) bits set.  These packets are retransmitted every
        RxmtInterval seconds.

        In state Exchange the Database Description Packets actually
        contain summaries of the link state information contained in the
        router's database.  Each LSA in the area's link-state database
        (at the time the neighbor transitions into Exchange state) is
        listed in the neighbor Database summary list.  Each new Database



Moy                         Standards Track                   [Page 103]

RFC 2328                     OSPF Version 2                   April 1998


        Description Packet copies its DD sequence number from the
        neighbor data structure and then describes the current top of
        the Database summary list.  Items are removed from the Database
        summary list when the previous packet is acknowledged.

        In state Exchange, the determination of when to send a Database
        Description packet depends on whether the router is master or
        slave:


        Master
            Database Description packets are sent when either a) the
            slave acknowledges the previous Database Description packet
            by echoing the DD sequence number or b) RxmtInterval seconds
            elapse without an acknowledgment, in which case the previous
            Database Description packet is retransmitted.

        Slave
            Database Description packets are sent only in response to
            Database Description packets received from the master.  If
            the Database Description packet received from the master is
            new, a new Database Description packet is sent, otherwise
            the previous Database Description packet is resent.


        In states Loading and Full the slave must resend its last
        Database Description packet in response to duplicate Database
        Description packets received from the master.  For this reason
        the slave must wait RouterDeadInterval seconds before freeing
        the last Database Description packet.  Reception of a Database
        Description packet from the master after this interval will
        generate a SeqNumberMismatch neighbor event.


    10.9.  Sending Link State Request Packets



        In neighbor states Exchange or Loading, the Link state request
        list contains a list of those LSAs that need to be obtained from
        the neighbor.  To request these LSAs, a router sends the
        neighbor the beginning of the Link state request list, packaged
        in a Link State Request packet.




Moy                         Standards Track                   [Page 104]

RFC 2328                     OSPF Version 2                   April 1998


        When the neighbor responds to these requests with the proper
        Link State Update packet(s), the Link state request list is
        truncated and a new Link State Request packet is sent.  This
        process continues until the Link state request list becomes
        empty. LSAs on the Link state request list that have been
        requested, but not yet received, are packaged into Link State
        Request packets for retransmission at intervals of RxmtInterval.
        There should be at most one Link State Request packet
        outstanding at any one time.

        When the Link state request list becomes empty, and the neighbor
        state is Loading (i.e., a complete sequence of Database
        Description packets has been sent to and received from the
        neighbor), the Loading Done neighbor event is generated.


    10.10.  An Example



        Figure 14 shows an example of an adjacency forming.  Routers RT1
        and RT2 are both connected to a broadcast network.  It is
        assumed that RT2 is the Designated Router for the network, and
        that RT2 has a higher Router ID than Router RT1.

        The neighbor state changes realized by each router are listed on
        the sides of the figure.

        At the beginning of Figure 14, Router RT1's interface to the
        network becomes operational.  It begins sending Hello Packets,
        although it doesn't know the identity of the Designated Router
        or of any other neighboring routers.  Router RT2 hears this
        hello (moving the neighbor to Init state), and in its next Hello
        Packet indicates that it is itself the Designated Router and
        that it has heard Hello Packets from RT1.  This in turn causes
        RT1 to go to state ExStart, as it starts to bring up the
        adjacency.

        RT1 begins by asserting itself as the master.  When it sees that
        RT2 is indeed the master (because of RT2's higher Router ID),
        RT1 transitions to slave state and adopts its neighbor's DD
        sequence number.  Database Description packets are then
        exchanged, with polls coming from the master (RT2) and responses
        from the slave (RT1).  This sequence of Database Description



Moy                         Standards Track                   [Page 105]

RFC 2328                     OSPF Version 2                   April 1998







            +---+                                         +---+
            |RT1|                                         |RT2|
            +---+                                         +---+

            Down                                          Down
                            Hello(DR=0,seen=0)
                       ------------------------------>
                         Hello (DR=RT2,seen=RT1,...)      Init
                       <------------------------------
            ExStart        D-D (Seq=x,I,M,Master)
                       ------------------------------>
                           D-D (Seq=y,I,M,Master)         ExStart
                       <------------------------------
            Exchange       D-D (Seq=y,M,Slave)
                       ------------------------------>
                           D-D (Seq=y+1,M,Master)         Exchange
                       <------------------------------
                           D-D (Seq=y+1,M,Slave)
                       ------------------------------>
                                     ...
                                     ...
                                     ...
                           D-D (Seq=y+n, Master)
                       <------------------------------
                           D-D (Seq=y+n, Slave)
             Loading   ------------------------------>
                                 LS Request                Full
                       ------------------------------>
                                 LS Update
                       <------------------------------
                                 LS Request
                       ------------------------------>
                                 LS Update
                       <------------------------------
             Full





Moy                         Standards Track                   [Page 106]

RFC 2328                     OSPF Version 2                   April 1998


                   Figure 14: An adjacency bring-up example





        Packets ends when both the poll and associated response has the
        M-bit off.

        In this example, it is assumed that RT2 has a completely up to
        date database.  In that case, RT2 goes immediately into Full
        state.  RT1 will go into Full state after updating the necessary
        parts of its database.  This is done by sending Link State
        Request Packets, and receiving Link State Update Packets in
        response.  Note that, while RT1 has waited until a complete set
        of Database Description Packets has been received (from RT2)
        before sending any Link State Request Packets, this need not be
        the case.  RT1 could have interleaved the sending of Link State
        Request Packets with the reception of Database Description
        Packets.


11.  The Routing Table Structure



    The routing table data structure contains all the information
    necessary to forward an IP data packet toward its destination.  Each
    routing table entry describes the collection of best paths to a
    particular destination.  When forwarding an IP data packet, the
    routing table entry providing the best match for the packet's IP
    destination is located.  The matching routing table entry then
    provides the next hop towards the packet's destination.  OSPF also
    provides for the existence of a default route (Destination ID =
    DefaultDestination, Address Mask =  0x00000000).  When the default
    route exists, it matches all IP destinations (although any other
    matching entry is a better match).  Finding the routing table entry
    that best matches an IP destination is further described in Section
    11.1.

    There is a single routing table in each router.  Two sample routing
    tables are described in Sections 11.2 and 11.3.  The building of the
    routing table is discussed in Section 16.




Moy                         Standards Track                   [Page 107]

RFC 2328                     OSPF Version 2                   April 1998


    The rest of this section defines the fields found in a routing table
    entry.  The first set of fields describes the routing table entry's
    destination.


    Destination Type
        Destination type is either "network" or "router". Only network
        entries are actually used when forwarding IP data traffic.
        Router routing table entries are used solely as intermediate
        steps in the routing table build process.

        A network is a range of IP addresses, to which IP data traffic
        may be forwarded.  This includes IP networks (class A, B, or C),
        IP subnets, IP supernets and single IP hosts.  The default route
        also falls into this category.

        Router entries are kept for area border routers and AS boundary
        routers.  Routing table entries for area border routers are used
        when calculating the inter-area routes (see Section 16.2), and
        when maintaining configured virtual links (see Section 15).
        Routing table entries for AS boundary routers are used when
        calculating the AS external routes (see Section 16.4).

    Destination ID
        The destination's identifier or name.  This depends on the
        Destination Type.  For networks, the identifier is their
        associated IP address.  For routers, the identifier is the OSPF
        Router ID.[9]

    Address Mask
        Only defined for networks.  The network's IP address together
        with its address mask defines a range of IP addresses.  For IP
        subnets, the address mask is referred to as the subnet mask.
        For host routes, the mask is "all ones" (0xffffffff).

    Optional Capabilities
        When the destination is a router this field indicates the
        optional OSPF capabilities supported by the destination router.
        The only optional capability defined by this specification is
        the ability to process AS-external-LSAs.  For a further
        discussion of OSPF's optional capabilities, see Section 4.5.




Moy                         Standards Track                   [Page 108]

RFC 2328                     OSPF Version 2                   April 1998


    The set of paths to use for a destination may vary based on the OSPF
    area to which the paths belong.  This means that there may be
    multiple routing table entries for the same destination, depending
    on the values of the next field.


    Area
        This field indicates the area whose link state information has
        led to the routing table entry's collection of paths.  This is
        called the entry's associated area.  For sets of AS external
        paths, this field is not defined.  For destinations of type
        "router", there may be separate sets of paths (and therefore
        separate routing table entries) associated with each of several
        areas. For example, this will happen when two area border
        routers share multiple areas in common.  For destinations of
        type "network", only the set of paths associated with the best
        area (the one providing the preferred route) is kept.


    The rest of the routing table entry describes the set of paths to
    the destination.  The following fields pertain to the set of paths
    as a whole.  In other words, each one of the paths contained in a
    routing table entry is of the same path-type and cost (see below).


    Path-type
        There are four possible types of paths used to route traffic to
        the destination, listed here in decreasing order of preference:
        intra-area, inter-area, type 1 external or type 2 external.
        Intra-area paths indicate destinations belonging to one of the
        router's attached areas.  Inter-area paths are paths to
        destinations in other OSPF areas.  These are discovered through
        the examination of received summary-LSAs.  AS external paths are
        paths to destinations external to the AS.  These are detected
        through the examination of received AS-external-LSAs.

    Cost
        The link state cost of the path to the destination.  For all
        paths except type 2 external paths this describes the entire
        path's cost.  For Type 2 external paths, this field describes
        the cost of the portion of the path internal to the AS.  This




Moy                         Standards Track                   [Page 109]

RFC 2328                     OSPF Version 2                   April 1998


        cost is calculated as the sum of the costs of the path's
        constituent links.

    Type 2 cost
        Only valid for type 2 external paths.  For these paths, this
        field indicates the cost of the path's external portion.  This
        cost has been advertised by an AS boundary router, and is the
        most significant part of the total path cost.  For example, a
        type 2 external path with type 2 cost of 5 is always preferred
        over a path with type 2 cost of 10, regardless of the cost of
        the two paths' internal components.

    Link State Origin
        Valid only for intra-area paths, this field indicates the LSA
        (router-LSA or network-LSA) that directly references the
        destination.  For example, if the destination is a transit
        network, this is the transit network's network-LSA.  If the
        destination is a stub network, this is the router-LSA for the
        attached router.  The LSA is discovered during the shortest-path
        tree calculation (see Section 16.1).  Multiple LSAs may
        reference the destination, however a tie-breaking scheme always
        reduces the choice to a single LSA. The Link State Origin field
        is not used by the OSPF protocol, but it is used by the routing
        table calculation in OSPF's Multicast routing extensions
        (MOSPF).

    When multiple paths of equal path-type and cost exist to a
    destination (called elsewhere "equal-cost" paths), they are stored
    in a single routing table entry.  Each one of the "equal-cost" paths
    is distinguished by the following fields:

    Next hop
        The outgoing router interface to use when forwarding traffic to
        the destination.  On broadcast, Point-to-MultiPoint and NBMA
        networks, the next hop also includes the IP address of the next
        router (if any) in the path towards the destination.

    Advertising router
        Valid only for inter-area and AS external paths.  This field
        indicates the Router ID of the router advertising the summary-
        LSA or AS-external-LSA that led to this path.




Moy                         Standards Track                   [Page 110]

RFC 2328                     OSPF Version 2                   April 1998


    11.1.  Routing table lookup



        When an IP data packet is received, an OSPF router finds the
        routing table entry that best matches the packet's destination.
        This routing table entry then provides the outgoing interface
        and next hop router to use in forwarding the packet. This
        section describes the process of finding the best matching
        routing table entry.

        Before the lookup begins, "discard" routing table entries should
        be inserted into the routing table for each of the router's
        active area address ranges (see Section 3.5).  (An area range is
        considered "active" if the range contains one or more networks
        reachable by intra-area paths.) The destination of a "discard"
        entry is the set of addresses described by its associated active
        area address range, and the path type of each "discard" entry is
        set to "inter-area".[10]

        Several routing table entries may match the destination address.
        In this case, the "best match" is the routing table entry that
        provides the most specific (longest) match. Another way of
        saying this is to choose the entry that specifies the narrowest
        range of IP addresses.[11] For example, the entry for the
        address/mask pair of (128.185.1.0, 0xffffff00) is more specific
        than an entry for the pair (128.185.0.0, 0xffff0000). The
        default route is the least specific match, since it matches all
        destinations. (Note that for any single routing table entry,
        multiple paths may be possible. In these cases, the calculations
        in Sections 16.1, 16.2, and 16.4 always yield the paths having
        the most preferential path-type, as described in Section 11).

        If there is no matching routing table entry, or the best match
        routing table entry is one of the above "discard" routing table
        entries, then the packet's IP destination is considered
        unreachable. Instead of being forwarded, the packet should then
        be discarded and an ICMP destination unreachable message should
        be returned to the packet's source.

    11.2.  Sample routing table, without areas



        Consider the Autonomous System pictured in Figure 2.  No OSPF
        areas have been configured.  A single metric is shown per



Moy                         Standards Track                   [Page 111]

RFC 2328                     OSPF Version 2                   April 1998


        outbound interface.  The calculation of Router RT6's routing
        table proceeds as described in Section 2.2.  The resulting
        routing table is shown in Table 12.  Destination types are
        abbreviated: Network as "N", Router as "R".

        There are no instances of multiple equal-cost shortest paths in
        this example.  Also, since there are no areas, there are no
        inter-area paths.

        Routers RT5 and RT7 are AS boundary routers.  Intra-area routes
        have been calculated to Routers RT5 and RT7.  This allows
        external routes to be calculated to the destinations advertised
        by RT5 and RT7 (i.e., Networks N12, N13, N14 and N15).  It is
        assumed all AS-external-LSAs originated by RT5 and RT7 are
        advertising type 1 external metrics.  This results in type 1
        external paths being calculated to destinations N12-N15.



    11.3.  Sample routing table, with areas



        Consider the previous example, this time split into OSPF areas.
        An OSPF area configuration is pictured in Figure 6.  Router
        RT4's routing table will be described for this area
        configuration.  Router RT4 has a connection to Area 1 and a
        backbone connection.  This causes Router RT4 to view the AS as
        the concatenation of the two graphs shown in Figures 7 and 8.
        The resulting routing table is displayed in Table 13.

        Again, Routers RT5 and RT7 are AS boundary routers.  Routers
        RT3, RT4, RT7, RT10 and RT11 are area border routers.  Note that
        there are two routing entries for the area border router RT3,
        since it has two areas in common with RT4 (Area 1 and the
        backbone).

        Backbone paths have been calculated to all area border routers.
        These are used when determining the inter-area routes.  Note
        that all of the inter-area routes are associated with the
        backbone; this is always the case when the calculating router is
        itself an area border router.  Routing information is condensed
        at area boundaries.  In this example, we assume that Area 3 has
        been defined so that networks N9-N11 and the host route to H1



Moy                         Standards Track                   [Page 112]

RFC 2328                     OSPF Version 2                   April 1998




      Type   Dest   Area   Path  Type    Cost   Next     Adv.
                                                Hop(s)   Router(s)
      ____________________________________________________________
      N      N1     0      intra-area    10     RT3      *
      N      N2     0      intra-area    10     RT3      *
      N      N3     0      intra-area    7      RT3      *
      N      N4     0      intra-area    8      RT3      *
      N      Ib     0      intra-area    7      *        *
      N      Ia     0      intra-area    12     RT10     *
      N      N6     0      intra-area    8      RT10     *
      N      N7     0      intra-area    12     RT10     *
      N      N8     0      intra-area    10     RT10     *
      N      N9     0      intra-area    11     RT10     *
      N      N10    0      intra-area    13     RT10     *
      N      N11    0      intra-area    14     RT10     *
      N      H1     0      intra-area    21     RT10     *
      R      RT5    0      intra-area    6      RT5      *
      R      RT7    0      intra-area    8      RT10     *
      ____________________________________________________________
      N      N12    *      type 1 ext.   10     RT10     RT7
      N      N13    *      type 1 ext.   14     RT5      RT5
      N      N14    *      type 1 ext.   14     RT5      RT5
      N      N15    *      type 1 ext.   17     RT10     RT7


               Table 12: The routing table for Router RT6
                         (no configured areas).

        are all condensed to a single route when advertised into the
        backbone (by Router RT11).  Note that the cost of this route is
        the maximum of the set of costs to its individual components.

        There is a virtual link configured between Routers RT10 and
        RT11.  Without this configured virtual link, RT11 would be
        unable to advertise a route for networks N9-N11 and Host H1 into
        the backbone, and there would not be an entry for these networks
        in Router RT4's routing table.

        In this example there are two equal-cost paths to Network N12.
        However, they both use the same next hop (Router RT5).



Moy                         Standards Track                   [Page 113]

RFC 2328                     OSPF Version 2                   April 1998


        Router RT4's routing table would improve (i.e., some of the
        paths in the routing table would become shorter) if an
        additional virtual link were configured between Router RT4 and
        Router RT3.  The new virtual link would itself be associated
        with the first entry for area border router RT3 in Table 13 (an
        intra-area path through Area 1).  This would yield a cost of 1
        for the virtual link.  The routing table entries changes that
        would be caused by the addition of this virtual link are shown


   Type   Dest        Area   Path  Type    Cost   Next      Adv.
                                                  Hops(s)   Router(s)
   __________________________________________________________________
   N      N1          1      intra-area    4      RT1       *
   N      N2          1      intra-area    4      RT2       *
   N      N3          1      intra-area    1      *         *
   N      N4          1      intra-area    3      RT3       *
   R      RT3         1      intra-area    1      *         *
   __________________________________________________________________
   N      Ib          0      intra-area    22     RT5       *
   N      Ia          0      intra-area    27     RT5       *
   R      RT3         0      intra-area    21     RT5       *
   R      RT5         0      intra-area    8      *         *
   R      RT7         0      intra-area    14     RT5       *
   R      RT10        0      intra-area    22     RT5       *
   R      RT11        0      intra-area    25     RT5       *
   __________________________________________________________________
   N      N6          0      inter-area    15     RT5       RT7
   N      N7          0      inter-area    19     RT5       RT7
   N      N8          0      inter-area    18     RT5       RT7
   N      N9-N11,H1   0      inter-area    36     RT5       RT11
   __________________________________________________________________
   N      N12         *      type 1 ext.   16     RT5       RT5,RT7
   N      N13         *      type 1 ext.   16     RT5       RT5
   N      N14         *      type 1 ext.   16     RT5       RT5
   N      N15         *      type 1 ext.   23     RT5       RT7


                  Table 13: Router RT4's routing table
                       in the presence of areas.





Moy                         Standards Track                   [Page 114]

RFC 2328                     OSPF Version 2                   April 1998


        in Table 14.



12.  Link State Advertisements (LSAs)



    Each router in the Autonomous System originates one or more link
    state advertisements (LSAs).  This memo defines five distinct types
    of LSAs, which are described in Section 4.3.  The collection of LSAs
    forms the link-state database.  Each separate type of LSA has a
    separate function.  Router-LSAs and network-LSAs describe how an
    area's routers and networks are interconnected.  Summary-LSAs
    provide a way of condensing an area's routing information.  AS-
    external-LSAs provide a way of transparently advertising
    externally-derived routing information throughout the Autonomous
    System.

    Each LSA begins with a standard 20-byte header.  This LSA header is
    discussed below.







    Type   Dest        Area   Path  Type   Cost   Next     Adv.
                                                  Hop(s)   Router(s)
    ________________________________________________________________
    N      Ib          0      intra-area   16     RT3      *
    N      Ia          0      intra-area   21     RT3      *
    R      RT3         0      intra-area   1      *        *
    R      RT10        0      intra-area   16     RT3      *
    R      RT11        0      intra-area   19     RT3      *
    ________________________________________________________________
    N      N9-N11,H1   0      inter-area   30     RT3      RT11


                  Table 14: Changes resulting from an
                        additional virtual link.





Moy                         Standards Track                   [Page 115]

RFC 2328                     OSPF Version 2                   April 1998


    12.1.  The LSA Header



        The LSA header contains the LS type, Link State ID and
        Advertising Router fields.  The combination of these three
        fields uniquely identifies the LSA.

        There may be several instances of an LSA present in the
        Autonomous System, all at the same time.  It must then be
        determined which instance is more recent.  This determination is
        made by examining the LS sequence, LS checksum and LS age
        fields.  These fields are also contained in the 20-byte LSA
        header.

        Several of the OSPF packet types list LSAs.  When the instance
        is not important, an LSA is referred to by its LS type, Link
        State ID and Advertising Router (see Link State Request
        Packets).  Otherwise, the LS sequence number, LS age and LS
        checksum fields must also be referenced.

        A detailed explanation of the fields contained in the LSA header
        follows.


        12.1.1.  LS age



            This field is the age of the LSA in seconds.  It should be
            processed as an unsigned 16-bit integer.  It is set to 0
            when the LSA is originated.  It must be incremented by
            InfTransDelay on every hop of the flooding procedure.  LSAs
            are also aged as they are held in each router's database.

            The age of an LSA is never incremented past MaxAge.  LSAs
            having age MaxAge are not used in the routing table
            calculation.  When an LSA's age first reaches MaxAge, it is
            reflooded.  An LSA of age MaxAge is finally flushed from the
            database when it is no longer needed to ensure database
            synchronization.  For more information on the aging of LSAs,
            consult Section 14.

            The LS age field is examined when a router receives two
            instances of an LSA, both having identical LS sequence
            numbers and LS checksums.  An instance of age MaxAge is then



Moy                         Standards Track                   [Page 116]

RFC 2328                     OSPF Version 2                   April 1998


            always accepted as most recent; this allows old LSAs to be
            flushed quickly from the routing domain.  Otherwise, if the
            ages differ by more than MaxAgeDiff, the instance having the
            smaller age is accepted as most recent.[12] See Section 13.1
            for more details.


        12.1.2.  Options



            The Options field in the LSA header indicates which optional
            capabilities are associated with the LSA.  OSPF's optional
            capabilities are described in Section 4.5.  One optional
            capability is defined by this specification, represented by
            the E-bit found in the Options field.  The unrecognized bits
            in the Options field should be set to zero.

            The E-bit represents OSPF's ExternalRoutingCapability.  This
            bit should be set in all LSAs associated with the backbone,
            and all LSAs associated with non-stub areas (see Section
            3.6).  It should also be set in all AS-external-LSAs.  It
            should be reset in all router-LSAs, network-LSAs and
            summary-LSAs associated with a stub area.  For all LSAs, the
            setting of the E-bit is for informational purposes only; it
            does not affect the routing table calculation.


        12.1.3.  LS type



            The LS type field dictates the format and function of the
            LSA.  LSAs of different types have different names (e.g.,
            router-LSAs or network-LSAs).  All LSA types defined by this
            memo, except the AS-external-LSAs (LS type = 5), are flooded
            throughout a single area only.  AS-external-LSAs are flooded
            throughout the entire Autonomous System, excepting stub
            areas (see Section 3.6).  Each separate LSA type is briefly
            described below in Table 15.

        12.1.4.  Link State ID



            This field identifies the piece of the routing domain that
            is being described by the LSA.  Depending on the LSA's LS
            type, the Link State ID takes on the values listed in Table



Moy                         Standards Track                   [Page 117]

RFC 2328                     OSPF Version 2                   April 1998




            LS Type   LSA description
            ________________________________________________
            1         These are the router-LSAs.
                      They describe the collected
                       states of the router's
                      interfaces. For more information,
                      consult Section 12.4.1.
            ________________________________________________
            2         These are the network-LSAs.
                      They describe the set of routers
                      attached to the network. For
                      more information, consult
                      Section 12.4.2.
            ________________________________________________
            3 or 4    These are the summary-LSAs.
                      They describe inter-area routes,
                      and enable the condensation of
                      routing information at area
                      borders. Originated by area border
                      routers, the Type 3 summary-LSAs
                      describe routes to networks while the
                      Type 4 summary-LSAs describe routes to
                      AS boundary routers.
            ________________________________________________
            5         These are the AS-external-LSAs.
                      Originated by AS boundary routers,
                      they describe routes
                      to destinations external to the
                      Autonomous System. A default route for
                      the Autonomous System can also be
                      described by an AS-external-LSA.


            Table 15: OSPF link state advertisements (LSAs).

            16.


            Actually, for Type 3 summary-LSAs (LS type = 3) and AS-
            external-LSAs (LS type = 5), the Link State ID may



Moy                         Standards Track                   [Page 118]

RFC 2328                     OSPF Version 2                   April 1998




            LS Type   Link State ID
            _______________________________________________
            1         The originating router's Router ID.
            2         The IP interface address of the
                      network's Designated Router.
            3         The destination network's IP address.
            4         The Router ID of the described AS
                      boundary router.
            5         The destination network's IP address.


                   Table 16: The LSA's Link State ID.

            additionally have one or more of the destination network's
            "host" bits set. For example, when originating an AS-
            external-LSA for the network 10.0.0.0 with mask of
            255.0.0.0, the Link State ID can be set to anything in the
            range 10.0.0.0 through 10.255.255.255 inclusive (although
            10.0.0.0 should be used whenever possible). The freedom to
            set certain host bits allows a router to originate separate
            LSAs for two networks having the same address but different
            masks. See Appendix E for details.

            When the LSA is describing a network (LS type = 2, 3 or 5),
            the network's IP address is easily derived by masking the
            Link State ID with the network/subnet mask contained in the
            body of the LSA.  When the LSA is describing a router (LS
            type = 1 or 4), the Link State ID is always the described
            router's OSPF Router ID.

            When an AS-external-LSA (LS Type = 5) is describing a
            default route, its Link State ID is set to
            DefaultDestination (0.0.0.0).


        12.1.5.  Advertising Router



            This field specifies the OSPF Router ID of the LSA's
            originator.  For router-LSAs, this field is identical to the
            Link State ID field.  Network-LSAs are originated by the



Moy                         Standards Track                   [Page 119]

RFC 2328                     OSPF Version 2                   April 1998


            network's Designated Router.  Summary-LSAs originated by
            area border routers.  AS-external-LSAs are originated by AS
            boundary routers.


        12.1.6.  LS sequence number



            The sequence number field is a signed 32-bit integer.  It is
            used to detect old and duplicate LSAs.  The space of
            sequence numbers is linearly ordered.  The larger the
            sequence number (when compared as signed 32-bit integers)
            the more recent the LSA.  To describe to sequence number
            space more precisely, let N refer in the discussion below to
            the constant 2**31.

            The sequence number -N (0x80000000) is reserved (and
            unused).  This leaves -N + 1 (0x80000001) as the smallest
            (and therefore oldest) sequence number; this sequence number
            is referred to as the constant InitialSequenceNumber. A
            router uses InitialSequenceNumber the first time it
            originates any LSA.  Afterwards, the LSA's sequence number
            is incremented each time the router originates a new
            instance of the LSA.  When an attempt is made to increment
            the sequence number past the maximum value of N - 1
            (0x7fffffff; also referred to as MaxSequenceNumber), the
            current instance of the LSA must first be flushed from the
            routing domain.  This is done by prematurely aging the LSA
            (see Section 14.1) and reflooding it.  As soon as this flood
            has been acknowledged by all adjacent neighbors, a new
            instance can be originated with sequence number of
            InitialSequenceNumber.

            The router may be forced to promote the sequence number of
            one of its LSAs when a more recent instance of the LSA is
            unexpectedly received during the flooding process.  This
            should be a rare event.  This may indicate that an out-of-
            date LSA, originated by the router itself before its last
            restart/reload, still exists in the Autonomous System.  For
            more information see Section 13.4.






Moy                         Standards Track                   [Page 120]

RFC 2328                     OSPF Version 2                   April 1998