Network Working Working Group R. Callon Request for Comments: 1195 Digital Equipment Corporation December 1990 Use of OSI IS-IS for Routing in TCP/IP and Dual Environments Status of this Memo This RFC specifies a protocol on the IAB Standards Track for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "IAB Official Protocol Standards" for the standardization state and status of this protocol. Distribution of this memo is unlimited. This RFC is available in both postscript and text versions. Where possible, use of the postscript version is recommended. For example, this text version may have figures which are less informative or missing. Abstract This RFC specifies an integrated routing protocol, based on the OSI Intra-Domain IS-IS Routing Protocol, which may be used as an interior gateway protocol (IGP) to support TCP/IP as well as OSI. This allows a single routing protocol to be used to support pure IP environments, pure OSI environments, and dual environments. This specification was developed by the IS-IS working group of the Internet Engineering Task Force. The OSI IS-IS protocol has reached a mature state, and is ready for implementation and operational use. The most recent version of the OSI IS-IS protocol is contained in ISO DP 10589 . The proposed standard for using IS-IS for support of TCP/IP will therefore make use of this version (with a minor bug correction, as discussed in Annex B). We expect that future versions of this proposed standard will upgrade to the final International Standard version of IS-IS when available. Comments should be sent to "firstname.lastname@example.org". Contents 1 Introduction: Overview of the Protocol 1.1 What the Integrated IS-IS offers 1.2 Overview of the ISO IS-IS Protocol 1.3 Overview of the Integrated IS-IS 1.4 Support of Mixed Routing Domains
1.5 Advantages of Using Integrated IS-IS 2 Symbols and Abbreviations 3 Subnetwork Independent Functions 3.1 Exchange of Routing Information 3.2 Hierarchical Abbreviation of IP Reachability Information 3.3 Addressing Routers in IS-IS Packets 3.4 External Links 3.5 Type of Service Routing 3.6 Multiple LSPs and SNPs 3.7 IP-Only Operation 3.8 Encapsulation 3.9 Authentication 3.10 Order of Preference of Routes / Dijkstra Computation 4 Subnetwork Dependent Functions 4.1 Link Demultiplexing 4.2 Multiple IP Addresses per Interface 4.3 LANs, Designated Routers, and Pseudonodes 4.4 Maintaining Router Adjacencies 4.5 Forwarding to Incompatible Routers 5 Structure and Encoding of PDUs 5.1 Overview of IS-IS PDUs 5.2 Overview of IP-Specific Information for IS-IS 5.3 Encoding of IP-Specific Fields in IS-IS PDUs 6 Security Considerations 7 Author's Address 8 References A Inter-Domain Routing Protocol Information A.1 Inter-Domain Information Type A.2 Encoding B Encoding of Sequence Number Packets B.1 Level 1 Complete Sequence Numbers PDU B.2 Level 2 Complete Sequence Numbers PDU B.3 Level 1 Partial Sequence Numbers PDU B.4 Level 2 Partial Sequence Numbers PDU C Dijkstra Calculation and Forwarding C.1 SPF Algorithm for IP and Dual Use C.2 Forwarding of IP packets
D Use of the Authentication Field D.1 Authentication Field in IS-IS packets D.2 Authentication Type 1 - Simple Password E Interaction of the Integrated IS-IS with Brouters E.1 The Problem E.2 Possible Solutions Figures 1 ISO Hierarchical Address Structure 2 An Example 3 Encoding of Variable Length Fields 1 Introduction: Overview of the Protocol The TCP/IP protocol suite has been growing in importance as a multi- vendor communications architecture. With the anticipated emergence of OSI, we expect coexistence of TCP/IP and OSI to continue for an extended period of time. There is a critical need for routers to support both IP traffic and OSI traffic in parallel. There are two main methods that are available for routing protocols to support dual OSI and IP routers. One method, known as "Ships in the Night", makes use of completely independent routing protocols for each of the two protocol suites. This specification presents an alternate approach, which makes use of a single integrated protocol for interior routing (i.e., for calculating routes within a routing domain) for both protocol suites. This integrated protocol design is based on the OSI Intra-domain IS- IS routing protocol , with IP-specific functions added. This RFC is considered a companion to the OSI IS-IS Routing spec, and will only describe the required additional features. By supporting both IP and OSI traffic, this integrated protocol design supports traffic to IP hosts, OSI end systems, and dual end systems. This approach is "integrated" in the sense that the IS-IS protocol can be used to support pure-IP environments, pure-OSI environments, and dual environments. In addition, this approach allows interconnection of dual (IP and OSI) routing domains with other dual domains, with IP-only domains, and with OSI-only domains. The protocol specified here is based on the work of the IETF IS-IS working group. 1.1 What the Integrated IS-IS offers The integrated IS-IS provides a single routing protocol which will
simultaneously provide an efficient routing protocol for TCP/IP, and for OSI. This design makes use of the OSI IS-IS routing protocol, augmented with IP-specific information. This design provides explicit support for IP subnetting, variable subnet masks, TOS-based routing, and external routing. There is provision for authentication information, including the use of passwords or other mechanisms. The precise form of authentication mechanisms (other than passwords) is outside of the scope of this document. Both OSI and IP packets are forwarded "as is" -- i.e., they are transmitted directly over the underlying link layer services without the need for mutual encapsulation. The integrated IS-IS is a dynamic routing protocol, based on the SPF (Dijkstra) routing algorithm. The protocol described in this specification allows for mixing of IP-only, OSI-only, and dual (IP and OSI) routers, as defined below. An IP-only IS-IS router (or "IP-only" router) is defined to be a router which: (i) Uses IS-IS as the routing protocol for IP, as specified in this report; and (ii) Does not otherwise support OSI protocols. For example, such routers would not be able to forward OSI CLNP packets. An OSI-only router is defined to be a router which uses IS-IS as the routing protocol for OSI, as specified in . Generally, OSI-only routers may be expected to conform to OSI standards, and may be implemented independent of this specification. A dual IS-IS router (or "dual" router) is defined to be a router which uses IS-IS as a single integrated routing protocol for both IP and OSI, as specified in this report. This approach does not change the way that IP packets are handled. IP-only and dual routers are required to conform to the requirements of Internet Gateways . The integrated IS-IS protocol described in this report outlines an Interior Gateway Protocol (IGP) which will provide routing within a TCP/IP routing domain (i.e., autonomous system). Other aspects of router functionality (e.g., operation of ICMP, ARP, EGP, etc.) are not affected by this proposal. Similarly, this approach does not change the way that OSI packets are handled. There will be no change at all to the contents nor to the handling of ISO 8473 Data packets and Error Reports, nor to ISO 9542 Redirects and ES Hellos. ISO 9542 IS Hellos transmitted on LANs are similarly unchanged. ISO 9542 IS Hellos transmitted on point-to-point links are unchanged except for the addition of IP-related information. Similarly, other OSI packets (specifically those involved in the IS-IS intra-domain routing protocol) remain unchanged
except for the addition of IP-related information. This approach makes use of the existing IS-IS packets, with IP- specific fields added. Specifically: (i) authentication information may be added to all IS-IS packets; (ii) the protocols supported by each router, as well as each router's IP addresses, are specified in ISO 9542 IS Hello, IS-IS Hello and Link State Packets; (iii) internally reachable IP addresses are specified in all Link State Packets; and (iv) externally reachable IP addresses, and external routing protocol information, may be specified in level 2 Link State Packets. The detailed encoding and interpretation of this in formation is specified in sections 3, 4, and 5 of this RFC. The protocol described in this report may be used to provide routing in an IP-only routing domain, in which all routers are IP-only. Similarly, this protocol may be used to provide routing in a pure dual domain, in which all routers are dual. Finally, this protocol may be used to provide routing in a mixed domain, in which some routers are IP-only, some routers are OSI-only, and some routers are dual. The specific topological restrictions which apply in this latter case are described in detail in section 1.4 ("Support of Mixed Routing Domains"). The use of IS-IS for support of pure OSI domains is specified in . This protocol specification does not constrain which network management protocol(s) may be used to manage IS-IS-based routers. Management information bases (MIBs) for managing IP-only, OSI-only, and dual routers, compatible with CMIP, CMOT, and/or SNMP, are the subject of a separate, companion document . 1.2 Overview of the ISO IS-IS Protocol The IS-IS Routing Protocol has been developed in ISO to provide routing for pure OSI environments. In particular, IS-IS is designed to work in conjunction with ISO 8473 (The ISO Connectionless Network Layer Protocol ), and ISO 9542 (The ISO End System to Intermediate System Protocol ). This section briefly describes the manner in which IS-IS is used to support pure OSI environments. Enhancements for support of IP and dual environments are specified elsewhere in this report. In IS-IS, the network is partitioned into "routing domains". The boundaries of routing domains are defined by network management, by setting some links to be "exterior links". If a link is marked as "exterior", no IS-IS routing messages are sent on that link. Currently, ISO does not have a standard for inter-domain routing (i.e., for routing between separate autonomous routing domains).
Instead, manual configuration is used. The link is statically configured with the set of address prefixes reachable via that link, and with the method by which they can be reached (such as the DTE address to be dialed to reach that address, or the fact that the DTE address should be extracted from the IDP portion of the ISO address). OSI IS-IS routing makes use of two-level hierarchical routing. A routing domain is partitioned into areas. Level 1 routers know the topology in their area, including all routers and end systems in their area. However, level 1 routers do not know the identity of routers or destinations outside of their area. Level 1 routers forward all traffic for destinations outside of their area to a level 2 router in their area. Similarly, level 2 routers know the level 2 topology, and know which addresses are reachable via each level 2 router. However, level 2 routers do not need to know the topology within any level 1 area, except to the extent that a level 2 router may also be a level 1 router within a single area. Only level 2 routers can exchange data packets or routing information directly with external routers located outside of the routing domains. +----------------------+-------------------------------+ | IDP | DSP | +----------------------+-------------------------------+ . . . . . . . . . +-----+----------------+----------+--------------+-----+ | AFI | IDI | HO-DSP | ID | SEL | +-----+----------------+----------+--------------+-----+ Figure 1 - ISO Hierarchical Address Structure As illustrated in figure 1, ISO addresses are subdivided into the Initial Domain Part (IDP), and the Domain Specific Part (DSP). The IDP is the part which is standardized by ISO, and specifies the format and authority responsible for assigning the rest of the address. The DSP is assigned by whatever addressing authority is specified by the IDP. The DSP is further subdivided into a "High Order Part of DSP" (HO-DSP), a system identifier (ID), and an NSAP selector (SEL). The HO-DSP may use any format desired by the authority which is identified by the IDP. Together, the combination of [IDP, HO-DSP] identify both the routing domain and the area within the routing domain. The combination of [IDP,HO-DSP] may therefore be referred to as the "Area Address". Usually, all nodes in an area have the same area address. However, sometimes an area might have multiple addresses. Motivations for
allowing this are: - It might be desirable to change the address of an area. The most graceful way of changing an area from having address A to having address B is to first allow it to have both addresses A and B, and then after all nodes in the area have been modified to recognize both addresses, then one by one the nodes can be modified to "forget" address A. - It might be desirable to merge areas A and B into one area. The method for accomplishing this is to, one by one, add knowledge of address B into the A partition, and similarly add knowledge of address A into the B partition. - It might be desirable to partition an area C into two areas, A and B (where "A" might equal "C", in which case this example becomes one of removing a portion of an area). This would be accomplished by first introducing knowledge of address A into the appropriate nodes (those destined to become area A), and knowledge of address B into the appropriate nodes, and then one by one removing knowledge of address C. Since OSI addressing explicitly identifies the area, it is very easy for level 1 routers to identify packets going to destinations outside of their area, which need to be forwarded to level 2 routers. In IS-IS, there are two types of routers: - Level 1 intermediate systems -- these nodes route based on the ID portion of the ISO address. They route within an area. They recognize, based on the destination address in a packet, whether the destination is within the area. If so, they route towards the destination. If not, they route to the nearest level 2 router. - Level 2 intermediate systems -- these nodes route based on the area address (i.e., on the combination of [IDP, HO-DSP]). They route towards areas, without regard to the internal structure of an area. A level 2 IS may also be a level 1 IS in one area. A level 1 router will have the area portion of its address manually configured. It will refuse to become a neighbor with a node whose area addresses do not overlap its area addresses. However, if level 1 router has area addresses A, B, and C, and a neighbor has area addresses B and D, then the level 1 router will accept the other node as a neighbor. A level 2 router will accept another level 2 router as a neighbor, regardless of area address. However, if the area addresses do not
overlap, the link would be considered by both routers to be "level 2 only", and only level 2 LSPs would flow on the link. External links (to other routing domains) must be from level 2 routers. IS-IS provides an optional partition repair function. In the unlikely case that a level 1 area become partitioned, this function, if implemented, allows the partition to be repaired via use of level 2 routes. IS-IS requires that the set of level 2 routers be connected. Should the level 2 backbone become partitioned, there is no provision for use of level 1 links to repair a level 2 partition. In unusual cases, a single level 2 router may lose connectivity to the level 2 backbone. In this case the level 2 router will indicate in its level 1 LSPs that it is not "attached", thereby allowing level 1 routers in the area to route traffic for outside of the domain to a different level 2 router. Level 1 routers therefore route traffic to destinations outside of their area only to level 2 routers which indicate in their level 1 LSPs that they are "attached". An end system may autoconfigure the area portion of its address by extracting the area portion of a neighboring router's address. If this is the case, then an endnode will always accept a router as a neighbor. Since the standard does not specify that the end system MUST autoconfigure its area address, an end system may be configured with an area address. In this case the end system would ignore router neighbors with non-matching area addresses. Special treatment is necessary for broadcast subnetworks, such as LANs. This solves two sets of issues: (i) In the absence of special treatment, each router on the subnetwork would announce a link to every other router on the subnetwork, resulting in n-squared links reported; (ii) Again, in the absence of special treatment, each router on the LAN would report the same identical list of end systems on the LAN, resulting in substantial duplication. These problems are avoided by use of a "pseudonode", which represents the LAN. Each router on the LAN reports that it has a link to the pseudonode (rather than reporting a link to every other router on the LAN). One of the routers on the LAN is elected "designated router". The designated router then sends out an LSP on behalf of the pseudonode, reporting links to all of the routers on the LAN. This reduces the potential n-squared links to n links. In addition, only the pseudonode LSP includes the list of end systems on the LAN, thereby eliminating the potential duplication (for further information on designated routers and pseudonodes, see ).
The IS-IS provides for optional Quality of Service (QOS) routing, based on throughput (the default metric), delay, expense, or residual error probability. This is described in greater detail in section 3.5, and in . 1.3 Overview of the Integrated IS-IS The integrated IS-IS allows a single routing protocol to be used to route both IP and OSI packets. This implies that the same two-level hierarchy will be used for both IP and OSI routing. Each area will be specified to be either IP-only (only IP traffic can be routed in that particular area), OSI-only (only OSI traffic can be routed in that area), or dual (both IP and OSI traffic can be routed in the area). This proposal does not allow for partial overlap of OSI and IP areas. For example, if one area is OSI-only, and an other area is IP-only, then it is not permissible to have some routers be in both areas. Similarly, a single backbone is used for the routing domain. There is no provision for independent OSI and IP backbones. Similarly, within an IP-only or dual area, the amount of knowledge maintained by routers about specific IP destinations will be as similar as possible as for OSI. For example, IP-capable level 1 routers will maintain the topology within the area, and will be able to route directly to IP destinations within the area. However, IP- capable level 1 routers will not maintain information about destinations outside of the area. Just as in normal OSI routing, traffic to destinations outside of the area will be forwarded to the nearest level 2 router. Since IP routes to subnets, rather than to specific end systems, IP routers will not need to keep nor distribute lists of IP host identifiers (note that routes to hosts can be announced by using a subnet mask of all ones). The IP address structure allows networks to be partitioned into subnets, and allows subnets to be recursively subdivided into smaller subnets. However, it is undesireable to require any specific relationship between IP subnet addresses and IS-IS areas. For example, in many cases, the dual routers may be installed into existing environments, which already have assigned IP and/or OSI addresses. In addition, even if IP addresses are not already pre- assigned, the address limitations of IP constrain what addresses may be assigned. We therefore will not require any specific relationship between IP addresses and the area structure. The IP addresses can be assigned completely independently of the OSI addresses and IS-IS area structure. As will be described in section 3.2 ("Hierarchical Abbreviation of IP Reachability Information"), greater efficiency and scaling of the routing algorithm can be achieved if there is some correspondence between the IP address assignment structure and the
area structure. Within an area, level 1 routers exchange link state packets which identify the IP addresses reachable by each router. Specifically, zero or more [IP address, subnet mask, metric] combinations may be included in each Link State Packet. Each level 1 router is manually configured with the [IP address, subnet mask, metric] combinations which are reachable on each interface. A level 1 router routes as follows: - If a specified destination address matches an [IP address, subnet mask, metric] reachable within the area, the packet is routed via level 1 routing. - If a specified destination address does not match any [IP address, subnet mask, metric] combination listed as reachable within the area, the packet is routed towards the nearest level 2 router. Flexible use of the limited IP address space is important in order to cope with the anticipated growth of IP environments. Thus an area (and by implication a routing domain) may simultaneously make use of a variety of different address masks for different subnets in the area (or domain). Generally, if a specified destination address matches more than one [IP address, subnet mask] pair, the more specific address is the one routed towards (the one with more "1" bits in the mask -- this is known as "best match" routing). Level 2 routers include in their level 2 LSPs a complete list of [IP address, subnet mask, metric] specifying all IP addresses reachable in their area. As described in section 3, this information may be obtained from a combination of the level 1 LSPs (obtained from level 1 routers in the same area), and/or by manual configuration. In addition, Level 2 routers may report external reachability information, corresponding to addresses which can be reached via routers in other routing domains (autonomous systems) Default routes may be announced by use of a subnet mask containing all zeroes. Default routes should be used with great care, since they can result in "black holes". Default routes are permitted only at level 2 as external routes (i.e., included in the "IP External Reachability Information" field, as explained in sections 3 and 5). Default routes are not permitted at level 1. The integrated IS-IS provides optional Type of Service (TOS) routing, through use of the QOS feature from IS-IS.
1.4 Support of Mixed Routing Domains The integrated IS-IS proposal specifically allows for three types of routing domains: - Pure IP - Pure OSI - Dual In a pure IP routing domain, all routers must be IP-capable. IP-only routers may be freely mixed with dual routers. Some fields specifically related to OSI operation may be included by dual routers, and will be ignored by IP-only routers. Only IP traffic will be routed in an pure IP domain. Any OSI traffic may be discarded (except for the IS-IS packets necessary for operation of the routing protocol). In a pure OSI routing domain, all routers must be OSI-capable. OSI- only routers may be freely mixed with dual routers. Some fields specifically related to IP operation may be included by dual routers, and will be ignored by OSI-only routers. Only OSI traffic will be routed in a pure OSI domain. Any IP traffic may be discarded. In a dual routing domain, IP-only, OSI-only, and dual routers may be mixed on a per-area basis. Specifically, each area may itself be defined to be pure IP, pure OSI, or dual. In a pure IP area within a dual domain, IP-only and dual routers may be freely mixed. Only IP traffic can be routed by level 1 routing within a pure-IP area. In a pure-OSI area within a dual domain, OSI-only and dual routers may be freely mixed. Only OSI traffic can be routed by level 1 routing within a pure OSI area. In a dual area within a dual routing domain only dual routers may be used. Both IP and OSI traffic can be routed within a dual area. Within a dual domain, if both IP and OSI traffic are to be routed between areas then all level 2 routers must be dual. 1.5 Advantages of Using Integrated IS-IS Use of the integrated IS-IS protocol, as a single protocol for routing both IP and OSI packets in a dual environment, has significant advantages over using separate protocols for
independently routing IP and OSI traffic. An alternative approach is known as "Ships In the Night" (S.I.N.). With the S.I.N. approach, completely separate routing protocols are used for IP and for OSI. For example, OSPF  may be used for routing IP traffic, and IS-IS  may be used for routing OSI traffic. With S.I.N., the two routing protocols operate more or less independently. However, dual routers will need to implement both routing protocols, and therefore there will be some degree of competition for resources. Note that S.I.N. and the integrated IS-IS approach are not really completely separate options. In particular, if the integrated IS-IS is used within a routing domain for routing of IP and OSI traffic, it is still possible to use other independent routing protocols for routing other protocol suites. In the future, optional extensions to IS-IS may be defined for routing other common protocol suites. However, such future options are outside of the scope of this document. This section will compare integrated IS-IS and S.I.N. for routing of IP and OSI only. A primary advantage of the integrated IS-IS relates to the network management effort required. Since the integrated IS-IS provides a single routing protocol, within a single coordinated routing domain using a single backbone, this implies that there is less information to configure. This combined with a single coordinated MIB simplifies network management. Note that the operation of two routing protocols with the S.I.N. approach are not really independent, since they must share common resources. However, with the integrated IS-IS, the interactions are explicit, whereas with S.I.N., the interactions are implicit. Since the interactions are explicit, again it may be easier to manage and debug dual routers. Another advantage of the integrated IS-IS is that, since it requires only one routing protocol, it uses fewer resources. In particular, less implementation resources are needed (since only one protocol needs to be implemented), less CPU and memory resources are used in the router (since only one protocol needs to be run), and less network resources are used (since only one set of routing packets need to be transmitted). Primarily this translates into a financial savings, since each of these three types of resources cost money. This implies that dual routers based on the integrated IS-IS should be less expensive to purchase and operate than dual routers based on S.I.N.
Note that the operation of two routing protocols with the S.I.N. approach are not really independent, since they must share common resources. For example, if one routing protocol becomes unstable and starts to use excessive resources, the other protocol is likely to suffer. A bug in one protocol could crash the other. However, with the integrated IS-IS, the interactions are explicit and are defined into the protocol and software interactions. With S.I.N., the interactions are implicit. The use of a single integrated routing protocol similarly reduces the likely frequency of software upgrades. Specifically, if you have two different routing protocols in your router, then you have to upgrade the software any time EITHER of the protocols change. If you make use of a single integrated routing protocol, then software changes are still likely to be needed, but less frequently. Finally, routing protocols have significant real time requirements. In IS-IS, these real time requirements have been explicitly specified. In other routing protocols, these requirements are implicit. However, in all routing protocols, there are real time guarantees which must be met in order to ensure correct operation. In general, it is difficult enough to ensure compliance with real time requirements in the implementation of a single real time system. With S.I.N., implementation of two semi-independent real-time protocols in a single device makes this more difficult. Note that both integrated IS-IS and S.I.N. allow for independence of external routes (for traffic from/to outside of the routing domain), and allow for independent assignment of OSI and TCP/IP addresses. 2 Symbols and Abbreviations AA Administrative Authority (a three octet field in the GOSIP version 2.0 NSAP address format) AFI Authority and Format Identifier (the first octet of all OSI NSAP addresses -- identifies format of the rest of the address) CLNP Connection-Less Network Protocol (ISO 8473, the OSI connectionless network layer protocol -- very similar to IP) DFI DSP Format Identifier (a one octet field in the GOSIP version 2.0 NSAP address format)
ES End System (The OSI term for a host) ES-IS End System to Intermediate System Routeing Exchange Protocol (ISO 9542 -- OSI protocol between routers and end systems) ICD International Code Designator (ISO standard for identifying organizations) IP Internetwork Protocol (an Internet Standard Network Layer Protocol) IS Intermediate System (The OSI term for a router) IS-IS Intermediate System to Intermediate System Routeing Exchange Protocol (the ISO protocol for routing within a single routing domain) IS-IS Hello An Hello packet defined by the IS-IS protocol (a type of packet used by the IS-IS protocol) ISH An Hello packet defined by ISO 9542 (ES-IS protocol). (not the same as IS-IS Hello) ISO International Organization for Standardization (an international body which is authorized to write standards of many kinds) LSP Link State Packet (a type of packet used by the IS-IS protocol) NLPID Network Layer Protocol ID (A one-octet field identifying a network layer protocol) NSAP Network Service Access Point (a conceptual interface point at which the network service is made available) SEL NSAP Selector (the last octet of NSAP addresses, also called NSEL) OSI Open Systems Interconnection (an international standard protocol architecture)
RD Routing Domain (the set of routers and end systems using a single instance of a routing protocol such as IS-IS) SNPA Subnetwork Point of Attachment (a conceptual interface at which a subnetwork service is provided) TCP Transmission Control Protocol (an Internet Standard Transport Layer Protocol) TCP/IP The protocol suite based on TCP, IP, and related protocols (the Internet standard protocol architecture) 3 Subnetwork Independent Functions 3.1 Exchange of Routing Information The exchange of routing information between routers makes use of the normal routing packet exchange as defined in the OSI IS-IS routing spec, with additional IP-specific information added to the IS-IS routing packets. The IS-IS protocol provides for the inclusion of variable length fields in all IS-IS packets. These fields are encoded using a "Code, Length, Value" triplet, where the code and length are encoded in one octet each, and the value has the length specified (from 0 to 254 octets). IS-IS requires that: "Any codes in a received PDU that are not recognised are ignored and passed through unchanged". This requirement applies to all routers implementing IS-IS, including OSI-only, IP-only, and dual routers. This allows IP-specific information to be encoded in a manner which OSI-only routers will ignore, and also allows OSI-specific information to be encoded in a manner which IP-only routers will ignore. IP-capable (i.e., all IP-only and dual) routers need to know what network layer protocols are supported by other routers in their area. This information is made available by inclusion of a "protocols supported" field in all IS-IS Hello and Link State Packets. This field makes use of the NLPID (Network Layer Protocol Identifier), which is a one-octet value assigned by ISO to identify network level protocols. NLPID values have been assigned to ISO 8473 and to IP. IP-capable routers need to know the IP address of the adjacent interface of neighboring routers. This is required for sending ICMP redirects (when an IP-capable router sends an ICMP redirect to a host, it must include the IP address of the appropriate interface of
the correct next-hop router). This information is made available by inclusion of the IP interface address in the IS-IS Hello packets. Specifically, each IS-IS Hello packet contains the IP address(es) of the interface over which the Hello is transmitted. The IS-IS allows multiple IP addresses to be assigned to each physical interface. In some cases, it will be useful for IP-capable routers to be able to determine an IP address(es) of all other routers at their level (i.e., for level 1 routers: all other routers in their area; for level 2 routers: all other level 2 routers in the routing domain). This is useful whenever an IP packet is to be sent to a router, such as for encapsulation or for transmission of network management packets. This information is made available by inclusion of IP address in LSPs. Specifically, each IS-IS LSP includes one or more IP addresses of the router which transmits the LSP. An IP-capable router is required to include at least one of its IP addresses in its LSPs, and may optionally include several or all of its IP addresses. Where a single router operates as both a level 1 and a level 2 router, it is required to include the same IP address(es) in its level 1 and level 2 LSPs. IP-capable routers need to know, for any given IP destination address, the correct route to that destination. Specifically, level 1 routers need to know what IP addresses are reachable from each level 1 router in their area. In addition, level 1 routers need to find level 2 routers (for traffic to IP addresses outside of their area). Level 2 routers need to know what IP addresses are reachable internally (either directly, or via level 1 routing) from other level 2 routers, and what addresses are reachable externally from other level 2 routers. All of this information is made available by inclusion of IP reachable address information in the Link State Packets. Internal (within the routing domain) and external (outside the domain) reachability information is announced separately in level 2 LSPs. Reachable IP addresses include a default metric, and may include multiple TOS-specific metrics. In general, for external routes, metrics may be of type "internal" (i.e., directly comparable with internal metrics) or of type "external" (i.e., not comparable with the internal metric). A route using internal metrics (i.e., either announced as "IP internal reachability information", or announced as "IP external reachability information" with an internal metric) is always preferred to a route using external metrics (i.e., announced as "IP external reachability information", with an external metric). The detailed encoding of the IP-specific information included in routing packets is provided in section 5 (Structure and Encoding of
PDUs). 3.2 Hierarchical Abbreviation of IP Reachability Information Level 2 routers include in their level 2 LSPs a list of all [IP address, subnet mask, metric] combinations reachable in their area. In general, this information may be determined from the level 1 LSPs from all routers in the area. If we ignore resource constraints, then it would be permissible for a level 2 router to simply duplicate all [IP address, subnet mask, metric] entries from all level 1 routers in its area (with appropriate metric adjustment), for inclusion in its level 2 LSP. However, in order for hierarchical routing to scale to large routing domain sizes, it is highly desired to abbreviate the reachable address information. This is accomplished by manual configuration of summary addresses. Each level 2 router may be configured with one or more [IP address, subnet mask, metric] entries for announcement in their level 2 LSPs. The set of reachable addresses obtained from level 1 LSPs is compared with the configured reachable addresses. Redundant information obtained from level 1 LSPs is not included in level 2 LSPs. Generally it is expected that the level 2 configured information will specify more inclusive addresses (corresponding to a subnet mask with fewer bits set to 1). This will therefore allow one configured address/submask pair (or a small number of such pairs) to hierarchically supercede the information corresponding to multiple entries in level 1 LSPs. The manually configured addresses are included in level 2 LSPs only if they correspond to at least one address which is reachable in the area. For manually configured level 2 addresses, the associated metric values to announce in level 2 LSPs are also manually configured. The configured addresses will supercede reachable address entries from level 1 LSPs based only on the IP address and subnet mask -- metric values are not considered when determining if a given configured address supercedes an address obtained from a level 1 LSP. Any address obtained from a level 1 LSP which is not superceded by the manually configured information is included in the level 2 LSPs. In this case, the metric value announced in the level 2 LSPs is calculated from the sum of the metric value announced in the corresponding level 1 LSP, plus the distance from the level 2 router to the appropriate level 1 router. Note: If this sum results in a metric value greater than 63 (the maximum value that can be reported in level 2 LSPs), then the value 63 must be used. Delay, expense, and error metrics (i.e., those TOS metrics other than the default metric) will be included only if (i) the level 2 router supports the specific
TOS; (ii) the path from the level 2 router to the appropropriate level 1 router is made up of links which support the specific TOS; and (iii) the level 1 router which can reach the address directly also supports the specific TOS for this route, as indicated in its level 1 LSP. In general, the same [IP address, subnet mask] pair may be announced in level 1 LSPs sent by multiple level 1 routers in the same area. In this case (assuming the entry is not superceded by a manually configured entry), then only one such entry shall be included in the level 2 LSP. The metric value(s) announced in level 2 LSPs correspond to the minimum of the metric value(s) that would be calculated for each of the level 1 LSP entries. A level 2 router will have IP addresses which are directly reachable via its own interfaces. For purposes of inclusion of IP reachable address information in level 2 LSPs, these "directly reachable" addresses are treated exactly the same as addresses received in level 1 LSPs. Manually configured addresses may hierarchically supercede multiple level 1 reachable address entries. However, there may be some IP addresses which match the manually configured addresses, but which are not reachable via level 1 routing. If a level 2 router receives an IP packet whose IP address matches a manually configured address which it is including in its level 2 LSP, but which is not reachable via level 1 routing in the area, then the packet must be discarded. In this case, an error report may be returned (as specified in RFC 1009), with the reason for discard specifying destination unreachable. Figure 2 - An Example Routing Domain (not shown) An example is illustrated in figure 2. Suppose that the network number for the entire routing domain is 17 (a class A network). Suppose each area is assigned a subnet number consisting of the next 8 bits. The area may be further subdivided by assigning the next eight bits to each LAN in the area, giving each a 24 bit subnet mask (counting the network and subnet fields). Finally 8 bits are left for the host field. Suppose that for a particular area (given subnet number 17.133) there are a number of IP capable level 1 routers announcing (in the special IP entry in their level 1 LSPs) subnets 17.133.5, 17.133.43, and 17.133.57.
Suppose that in this example, in order to save space in level 2 LSPs, the level 2 routers in this area are configured to announce subnet 17.133. Only this one address needs to be announced in level 2 LSPs. Thus if an IP packet comes along for an address in subnet 17.133.5, 17.133.43 or 17.133.57, then other level 2 routers, in other areas, will know to pass the traffic to this area. The inclusion of 17.133 in level 2 LSPs means that the three subnet addresses starting with 17.133 do not all have to be listed separately in level 2 LSPs. If any traffic comes along that is for an unreachable address such as 184.108.40.206, then level 2 routers in other areas in this particular domain will think that this area can handle this traffic, will forward traffic to level 2 routers in this area, which will have to discard this traffic. Suppose that subnet number 17.133.125 was actually reachable via some other area, such as the lower right hand area. In this case, the level 2 router in the left area would be announcing (in its level 2 LSPs according to manually configured information) reachability to subnet 17.133. However, the level 2 router in the lower right area would be announcing (in its level 2 LSPs according to information taken from its received level 1 LSPs), reachability to subnet 17.133.125. Due to the use of best match routing, this works correctly. All traffic from other areas destined to subnet 17.133.125 would be sent to the level 2 router in the lower right area, and all other traffic to subnet 17.133 (i.e., traffic to any IP address starting with 17.133, but not starting with 17.133.125) would be sent to the level 2 router in the leftmost area. 3.3 Addressing Routers in IS-IS Packets The IS-IS packet formats explicitly require that OSI-style addresses of routers appear in the IS-IS packets. For example, these addresses are used to determine area membership of routers. It is therefore necessary for all routers making use of the IS-IS protocol to have OSI style addresses assigned. For IP-only routers, these addresses will be used only in the operation of the IS-IS protocol, and are not used for any other purpose (such as the operation of EGP, ICMP, or other TCP/IP protocols). For OSI-only and dual routers, assignment of NSAP addresses is straight forward, but is outside of the scope of this specification. Address assignment mechanisms are being set up by standards bodies which allow globally unique OSI NSAP addresses to be assigned. All OSI-only and dual routers may therefore make use of normal OSI addresses in the operation of the IS-IS protocol.
For IP-only routers, there are two ways in which NSAP addresses may be obtained for use with the IS-IS protocol. 1) For those environments in which OSI is being used, or in which it is anticipated that OSI will be used in the future, it is permissible to obtain NSAP address assignments in the normal manner, assign normal NSAP addresses to IP-only routers, and use these addresses in the operation of IS-IS. This approach is recommended even for pure IP routing domains, as it will simplify future migration from IP-only to dual operation. 2) In some cases, routers may have only TCP/IP addresses, and it may be undesireable to have to go through the normal mechanisms for assignment of NSAP addresses. Instead, an alternate mechanim is provided below for algorithmically generating a valid OSI style address from existing IP address and autonomous system number assignments. Where desired, for IP-only routers, for use in IS-IS packet formats only, OSI-style addresses (compatible with the USA GOSIP version 2.0 NSAP address format ) may be derived as follows: AFI 1 octet value "47" (specifies ICD format) ICD 2 octet value "00 05" (specifies Internet/Gosip) DFI 1 octet value "xx" AA 3 octets value "xx xx xx" (specifies special IP-only use of NSAPs) Reserved 2 octets must be "00 00" RD 2 octets contains autonomous system number Area 2 octets must be assigned as described below ID 6 octets must be assigned as described below SEL 1 octet used as described below The AFI value of "47" and the ICD value of "00 05" specifies the Gosip Version 2.0 addressing format. The DFI number of "xx" and the AA of "xx xx xx" specify that this special NSAP address format is being used, solely for IS-IS packet formats in an IP-only environment. The reserved field must contain "00 00", as specified in GOSIP version 2.0.
The routing domain field contains the Autonomous System number. Strictly speaking, this is not necessary, since the IS-IS packets are exchanged within a single AS only. However, inclusion of the AS number in this address format will ensure correct operation in the event that routers from separate routing domains/ASs are incorrectly placed on the same link. The AS number in this context is used only for definition of unique NSAP addresses, and does not imply any coupling with exterior routing protocols. The Area field must be assigned by the authority responsible for the routing domain, such that each area in the routing domain must have a unique Area value. The ID must be assigned by the authority responsible for the routing domain. The ID must be assigned such that every router in the routing domain has a unique value. It is recommended that one of the following methods is used: 1)use a unique IEEE 802 48 bit station ID 2)use the value hex "02 00" prepended to an IP address of the router. IEEE 802 addresses, if used, must appear in IEEE canonical format. Since the IEEE 802 station IDs are assigned to be globally unique, use of these values clearly assures uniqueness in the area. Also, all assigned IEEE 802 station IDs have the global/local bit set to zero. Prepending the indicated pattern to the front of the IP address therefore assures that format (2) illustrated above cannot produce addresses which collide with format (1). Finally, to the extent that IP addresses are also globally unique, format (2) will produce unique IDs for routers. The indicated hex value is specified in IEEE 802 canonical form . In IEEE 802 addresses, the multicast bit is the least significant bit of the first byte. The global/local bit is the next least significant bit of the first byte. The indicated prefix therefore sets the global/local bit to 1, and all other bits in the first two octets to 0. Note that within an area, whether ISO addresses are configured into the routers through ISO address assignment, or whether the ISO-style address is generated directly from the AS number and IP address, all routers within an area must have the same high order part of address (AFI, ICD, DFI, AA, RD, and Area). This ISO-style address is used in IS-IS Hello messages and is the basis by which routers recognize whether neighbor nodes are in or out of their area.
3.4 External Links External connectivity (i.e., communications with routers outside of the routing domain) is done only by level 2 routers. The ISO version of IS-IS allows external OSI routes to be reported as "reachable address prefixes" in level 2 LSPs. The integrated IS-IS also allows external IP reachable addresses (i.e., IP addresses reachable via inter-domain routing) to be reported in level 2 LSPs in the "IP external reachability information" field. External OSI and external IP routes are handled independently. The routes announced in IP external reachability information entries include all routes to outside of the routing domain. This includes routes learned from OSPF, EGP, RIP, or any other external protocol. External routes may make use of "internal" or "external" metrics. Internal metrics are comparable with the metrics used for internal routes. Thus in choosing between an internal route, and an external route using internal metrics, the metric values may be directly compared. In contrast, external metrics cannot be directly compared with internal metrics. Any route defined solely using internal metrics is always preferred to any route defined using external metrics. When an external route using external metrics must be used, the lowest value of the external metric is preferred regardless of the internal cost to reach the appropriate exit point. It is useful, in the operation of external routing protocols, to provide a mechanism for border routers (i.e., routers in the same routing domain, which have the ability to route externally to other domains) to determine each other's existence, and to exchange external information (in a form understood only by the border routers themselves). This is made possible by inclusion of "inter-domain routing protocol information" fields in level 2 LSPs. The inter- domain routing protocol information field is not included in pseudonode LSPs. In general there may be multiple types of external inter-domain routing protocol information exchanged between border routers. The IS-IS therefore specifies that each occurance of the inter-domain routing protocol information field include a "type" field, which indicates the type of inter-domain routing protocol information enclosed. Values to be used in the type field will be specified in future versions of the "Assigned Numbers" RFC. Initial values for this field are specified in Annex A of this specification. Information contained in the inter-domain routing protocol information field will be carried in level 2 LSPs, and will therefore need to be stored by all level 2 routers in the domain. However, only
those level 2 routers which are directly involved in external routing will use this information. In designing the use of this field, it is important to carefully consider the implications that this may have on storage requirements in level 2 routers (including those level 2 routers which are not directly involved in external routing). The protocols used to exchange routing information directly between border routers, and external routers (in other routing domains / autonomous systems) are outside of the scope of this specification. 3.5 Type of Service Routing The integrated IS-IS protocol provides IP Type of Service (TOS) routing, through use of the Quality of Service (QOS) feature of IS- IS. This allows for routing on the basis of throughput (the default metric), delay, expense, or residual error probability. Note than any particular packet may be routed on the basis of any one of these four metrics. Routing on the basis of general combinations of metrics is not supported. The support for TOS/QOS is optional. If a particular packet calls for a specific TOS, and the correct path from the source to destination is made up of routers all of which support that particular TOS, then the packet will be routed on the optimal path. However, if there is no path from the source to destination made up of routers which support that particular type of service, then the packet will be forwarded using the default metric instead. This allows for TOS service in those environments where it is needed, while still providing acceptable service in the case where an unsupported TOS is requested. NOTE - IP does not have a cost TOS. There is therefore no mapping of IP TOS metrics which corresponds to the minimum cost metric. The IP TOS field is mapped onto the four available metrics as follows: Bits 0-2 (Precedence): This field does not affect the route, but rather may affect other aspects of packet forwarding. Bits 3 (Delay), 4 (Throughput) and 5 (Reliability): 000 (all normal) Use default metric 100 (low delay) Use delay metric 010 (high throughput) Use default metric
001 (high reliabiity) Use reliability metric other Use default metric 3.6 Multiple LSPs and SNPs In some cases, IS-IS packets (specifically Link State Packets and Complete Sequence Number Packets) may be too large to fit into one packet. The OSI IS-IS  allows for LSPs and CSNPs to be split into multiple packets. This is independent of ISO 8473 segmentation, and is also independent of IP fragmentation. Use of independent multiple packets has the advantages (with respect to segmentation or fragmentation) that: (i) when information in the IS-IS changes, only those packets effected need to be re-issued; (ii) when a single packet is received, it can be processed without the need to receive all other packets of the same type from the same router before beginning processing. The Integrated IS-IS makes use of the same multiple packet function, as defined in . IP-specific fields in IS-IS packets may be split across multiple packets. As specified in section 5 ("Structure and Encoding of PDUs"), some of the IP-specific fields (those which may be fairly long) may be split into several occurences of the same field, thereby allowing splitting of the fields across different packets. Multiple LSPs from the same router are distinguished by LSP number. Generally, most variable length fields may occur in an LSP with any LSP number. Some specific variable length fields may be required to occur in LSP number 0. Except where explicitly stated otherwise, when an IS-IS router issues multiple LSPs, the IP-specific fields may occur in an LSP with any LSP number. Complete Sequence Number Packets may be split into multiple packets, with the range to which each packet applies explicitly reported in the packet. Partial Sequence Number Packets are inherently partial, and so can easily be split into multiple packets if this is necessary. Again, where applicable, IP-specific fields may occur in any SNP. 3.7 IP-Only Operation For IP-only routers, the format for IS-IS packets remains unchanged. However, there are some variable length fields from the IS-IS packets that can be omitted. Specifically:
IS-IS Hello Packets: - no change IS-IS Link State Packets: - the "End Systems Neighbours" entries are omitted - the "Prefix Neighbours" entries are omitted IS-IS Sequence Number Packets: - no change 3.8 Encapsulation Future versions of the Integated IS-IS may specify optional encapsulation mechanisms for partition repair, and for forwarding packets through incompatible routers (i.e., for forwarding OSI packets through IP-only routers, and forwarding IP packets through OSI-only routers). The details of encapsulation and decapsulation are for further study. Routers complying with the Integrated IS-IS are not required to implement encapsulation nor decapsulation. 3.9 Authentication The authentication field allows each IS-IS packet to contain information used to authenticate the originator and/or contents of the packet. The authentication information contained in each packet is used to authenticate the entire packet, including OSI and IP parts. If a packet is received which contains invalid authentication information, then the entire packet is discarded. If an LSP or SNP is split into multiple packets (as described in section 3.6), then each is authenticated independently. Use of the authentication field is optional. Routers are not required to be able to interpret authentication information. As with other fields in the integrated IS-IS, if a router does not implement authentication then it will ignore any authentication field that may be present in an IS-IS packet. Annex D specifies a proposed use of the authentication field. 3.10 Order of Preference of Routes / Dijkstra Computation We define the term "IP reachability entry" to mean the combination of the [IP address, subnet mask]. The Dijkstra calculation must calculate routes to each distinct IP reachability entry. For the
Dijkstra calculation, each IP reachability entry can be treated in much the same manner as an OSI end system. Naturally, each IP reachability entry is treated as distinct from any OSI end systems which may also be reachable in the same area or routing domain. For any particular IP reachability entry, this is the same as another entry if and only if: (i) the subnet masks are identical; and (ii) for each bit in the subnet mask which has the value "1", the IP address is identical. This can easily be tested by zeroing those bits in the IP address which correspond to a zero bit in the mask, and then treating the entry as a 64 bit quantity, and testing for equality between different 64 bit quantities. The actual calculation of routes to IP reachability entries is therefore no more complex than calculation of routes to OSI end systems (except for the replacement of a 48-bit test with a 64-bit test). The Dijkstra computation does not take into consideration whether a router is IP-only, OSI-only, or dual. The topological restrictions specified in section 1.4 ensure that IP packets will only be sent via IP-capable routers, and OSI packets will only be sent via OSI-capable routers. The Integrated IS-IS prefers routes within the area (via level 1 routing) whenever possible. If level 2 routes must be used, then routes within the routing domain (specifically, those routes using internal metrics) are prefered to routes outside of the routing domain (using external metrics). The Integrated IS-IS protocol makes use of "best match" routing of IP packets. This implies that a particular destination address may match more than one entry in the forwarding database. If a particular IP packet has a destination address which matches two different IP reachability entries, then the entry who's mask contains the most "1" bits is preferred. IP packets whose destination is a router are routed the same way as any other IP packet, by forwarding first to the appropriate subnet, and then forwarding on that subnet to the destination host (which just happens to be a router in this case). In particular, the IP forwarding database does not contain explicit routes to the individual "IP interface addresses" listed by each router in its LSP. However, host routes (routes with a subnet mask of all ones) may of course be included in the IP reachability entries, and will be handled in the same manner as other IP reachability entries. In order to ensure correct interoperation of different router implementations, it is necessary to specify the order of preference
of possible routes. For OSI destinations, this is outside of the scope of this report. For IP destinations, this is specified in section 3.10.1 and 3.10.2 below. Annex C specifies a detailed Dijkstra calculation and forwarding algorithm which is compatible with the order of preference of routes specified here. With IS-IS, if a route to a given destination is advertised, or a link between routers is advertised, then metric values associated with some or all of the specified TOS metric types may be associated with that destination or link. However, the default metric must always be available. Normally this ensures that if a route using any TOS metric is available, then a route using the default metric will also be available. The only exception to this is where the corresponding route using the default metric has a total cost (within the area, or within the level 2 backbone) greater than MaxPathMetric. In determining the route to a particular destination for a specified TOS, only routes using either the requested TOS metric, or the default TOS metric, are considered. 3.10.1 Order of Preference of Routes In Level 1 Routing If a given destination is reachable within an area via a route using either the requested TOS or the default TOS, then the IS-IS will always make use of a path within the area (via level 1 routing), regardless of whether an alternate path exists outside of the area (via level 2 routing). In this case, routes within the area are selected as follows: 1) Amongst routes in the area, if the specified destination address matches more than one [IP address, subnet mask] pair, then the more specific address match (the one with more "1" bits in the mask) is prefered. 2) Amongst routes in the area to equally specific address matches, routes on which the requested TOS (if any) is supported are always prefered to routes on which the requested TOS is not supported. 3) Amongst routes in the area of the same TOS to equally specific address matches, the shortest routes are prefered. For determination of the shortest path, if a route on which the specified TOS is supported is available, then the specified TOS metric is used, otherwise the default metric is used. Amongst routes of equal cost, load splitting may be performed as specified in . For a level 1 only router (i.e., a router which does not take part in
level 2 routing, or a level 2 router which is not "attached"), if a given destination is not reachable within an area, level 1 routing will always route to a level 2 router as follows: 1) Amongst routes in the area to attached level 2 routers, routes on which the requested TOS (if any) is supported are always prefered to routes on which the requested TOS is not supported. 2) Amongst routes in the area of the same TOS to attached level 2 routers, the shortest routes are prefered. For determination of the shortest path, if a route on which the specified TOS is supported is available, then the specified TOS metric is used, otherwise the default metric is used. Amongst routes of equal cost, loadsplitting may be performed as specified in . 3.10.2 Order of Preference of Routes in Level 2 Routing For those level 2 routers which also take part in level 1 routing, routes learned via level 1 routing, using either the requested TOS or the default TOS, are always prefered to routes learned through level 2 routing. For destinations which are not reachable via level 1 routing, or for level 2 only routers (routers which do not take part in level 1 routing), then level 2 routes are selected as follows: 1) Routes using internal metrics only are always preferred to routes using external metrics. 2) If a route using internal metrics only is available: a) If the specified destination address matches more than one [IP address, subnet mask] pair, then the more specific address match (i.e., the largest number of "1"s present in the subnet mask) is prefered. b) Amongst routes with equally specific address matches (i.e., an equal number of "1"s present in the subnet mask), routes on which the requested TOS (if any) is supported are always preferred to routes on which the requested TOS is not supported. c) Amongst routes of the same TOS with an equally specific address matches, the shortest path is prefered. For determination of the shortest path, if a route on which the specified TOS is supported is available, then the specified TOS metric is used, otherwise the default metric is used. Amongst routes of equal cost,
loadsplitting may be performed as specified in . NOTE: Internal routes (routes to destinations announced in the "IP Internal Reachability Information" field), and external routes using internal metrics (routes to destinations announced in the "IP External Reachability Information" field, with a metric of type "internal") are treated identically for the purpose of the order of preference of routes, and the Dijkstra calculation. 3) If a route using internal metrics only is not available, but a route using external metrics is available: a) If the specified destination address matches more than one [IP address, subnet mask] pair, then the more specific address match is prefered. NOTE: For external routes, the subnet mask will normally correspond precisely to the network number. This implies that this test will always discover equal length matching strings. However, this test is included to allow future migration to more general handling of external addresses. b) Amongst routes with equally specific matches, routes on which the requested TOS (if any) is supported are always preferred to routes on which the requested TOS is not supported. NOTE: for external routes, the route is considered to support the requested TOS only if the internal route to the appropriate border router supports the requested TOS, and the external route reported by the border router also supports the requested TOS. c) Amongst routes of the same TOS with an equal length matching address string, the shortest path is prefered. For determination of the shortest path: (i) Routes with a smaller announced external metric are always prefered. (ii) Amongst routes with an equal external metric, routes with a shorter internal metric are prefered. Amongst routes of equal cost, loadsplitting may be performed as specified in . For level 2 routers which are announcing manually configured summary addresses in their level 2 LSPs, in some cases there will exist IP addresses which match the manually configured addresses, but which do
not match any addresses which are actually reachable via level 1 routing in the area. Generally, packets to such addresses are handled according to the following rules: 1) If the specified destination is reachable via level 1 routing, then according to the order of preference of routes specified above, the packet will be delivered via level 1 routing. 2) If the specified destination is not reachable via level 1 routing, but is reachable via 2 routing, and there are other level 2 routers which offer more desireable routes according to the rules specified above (for example a route with a more specific match, or a route with an equally specific match which supports the correct TOS), then level 2 routing will forward the packet according to the more desireable route. 3) If the specified destination is not reachable via level 1 routing, and the manually configured summary address advertised by this router (the router which has received the packet and is trying to forward it) represents the most desireable route, then the destination is unreachable and the packet must be discarded.