RFC 1349
This document is obsolete. Please refer to RFC 2474.





Network Working Group                                        P. Almquist
Request for Comments: 1349                                    Consultant
Updates: RFCs 1248, 1247, 1195,                                July 1992
         1123, 1122, 1060, 791




             Type of Service in the Internet Protocol Suite

Status of This Memo



   This document specifies an IAB standards track protocol 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.

Summary

   This memo changes and clarifies some aspects of the semantics of the
   Type of Service octet in the Internet Protocol (IP) header.  The
   handling of IP Type of Service by both hosts and routers is specified
   in some detail.

   This memo defines a new TOS value for requesting that the network
   minimize the monetary cost of transmitting a datagram.  A number of
   additional new TOS values are reserved for future experimentation and
   standardization.  The ability to request that transmission be
   optimized along multiple axes (previously accomplished by setting
   multiple TOS bits simultaneously) is removed.  Thus, for example, a
   single datagram can no longer request that the network simultaneously
   minimize delay and maximize throughput.

   In addition, there is a minor conflict between the Host Requirements
   (RFC-1122 and RFC-1123) and a number of other standards concerning
   the sizes of the fields in the Type of Service octet.  This memo
   resolves that conflict.

Table of Contents



   1.  Introduction ...............................................    3

   2.  Goals and Philosophy .......................................    3

   3.  Specification of the Type of Service Octet .................    4

   4.  Specification of the TOS Field .............................    5



Almquist                                                        [Page 1]



RFC 1349                    Type of Service                    July 1992


   5.  Use of the TOS Field in the Internet Protocols .............    6
      5.1  Internet Control Message Protocol (ICMP) ...............    6
      5.2  Transport Protocols ....................................    7
      5.3  Application Protocols ..................................    7

   6.  ICMP and the TOS Facility ..................................    8
      6.1  Destination Unreachable ................................    8
      6.2  Redirect ...............................................    9

   7.  Use of the TOS Field in Routing ............................    9
      7.1  Host Routing ...........................................   10
      7.2  Forwarding .............................................   12

   8.  Other consequences of TOS ..................................   13

   APPENDIX A.  Updates to Other Specifications ...................   14
      A.1  RFC-792 (ICMP) .........................................   14
      A.2  RFC-1060 (Assigned Numbers) ............................   14
      A.3  RFC-1122 and RFC-1123 (Host Requirements) ..............   16
      A.4  RFC-1195 (Integrated IS-IS) ............................   16
      A.5  RFC-1247 (OSPF) and RFC-1248 (OSPF MIB) ................   17

   APPENDIX B.  Rationale .........................................   18
      B.1  The Minimize Monetary Cost TOS Value ...................   18
      B.2  The Specification of the TOS Field .....................   19
      B.3  The Choice of Weak TOS Routing .........................   21
      B.4  The Retention of Longest Match Routing .................   22
      B.5  The Use of Destination Unreachable .....................   23

   APPENDIX C.  Limitations of the TOS Mechanism ..................   24
      C.1  Inherent Limitations ...................................   24
      C.2  Limitations of this Specification ......................   25

   References .....................................................   27

   Acknowledgements ...............................................   28

   Security Considerations ........................................   28

   Author's Address ...............................................   28











Almquist                                                        [Page 2]



RFC 1349                    Type of Service                    July 1992


1.  Introduction



   Paths through the Internet vary widely in the quality of service they
   provide.  Some paths are more reliable than others.  Some impose high
   call setup or per-packet charges, while others do not do usage-based
   charging.  Throughput and delay also vary widely.  Often there are
   tradeoffs: the path that provides the highest throughput may well not
   be the one that provides the lowest delay or the lowest monetary
   cost.  Therefore, the "optimal" path for a packet to follow through
   the Internet may depend on the needs of the application and its user.

   Because the Internet itself has no direct knowledge of how to
   optimize the path for a particular application or user, the IP
   protocol [11] provides a (rather limited) facility for upper layer
   protocols to convey hints to the Internet Layer about how the
   tradeoffs should be made for the particular packet.  This facility is
   the "Type of Service" facility, abbreviated as the "TOS facility" in
   this memo.

   Although the TOS facility has been a part of the IP specification
   since the beginning, it has been little used in the past.  However,
   the Internet host specification [1,2] now mandates that hosts use the
   TOS facility.  Additionally, routing protocols (including OSPF [10]
   and Integrated IS-IS [7]) have been developed which can compute
   routes separately for each type of service.  These new routing
   protocols make it practical for routers to consider the requested
   type of service when making routing decisions.

   This specification defines in detail how hosts and routers use the
   TOS facility.  Section 2 introduces the primary considerations that
   motivated the design choices in this specification.  Sections 3 and 4
   describe the Type of Service octet in the IP header and the values
   which the TOS field of that octet may contain.  Section 5 describes
   how a host (or router) chooses appropriate values to insert into the
   TOS fields of the IP datagrams it originates.  Sections 6 and 7
   describe the ICMP Destination Unreachable and Redirect messages and
   how TOS affects path choice by both hosts and routers.  Section 8
   describes some additional ways in which TOS may optionally affect
   packet processing.  Appendix A describes how this specification
   updates a number of existing specifications.  Appendices B and C
   expand on the discussion in Section 2.

2.  Goals and Philosophy



   The fundamental rule that guided this specification is that a host
   should never be penalized for using the TOS facility.  If a host
   makes appropriate use of the TOS facility, its network service should
   be at least as good as (and hopefully better than) it would have been



Almquist                                                        [Page 3]



RFC 1349                    Type of Service                    July 1992


   if the host had not used the facility.  This goal was considered
   particularly important because it is unlikely that any specification
   which did not meet this goal, no matter how good it might be in other
   respects, would ever become widely deployed and used.  A particular
   consequence of this goal is that if a network cannot provide the TOS
   requested in a packet, the network does not discard the packet but
   instead delivers it the same way it would have been delivered had
   none of the TOS bits been set.

   Even though the TOS facility has not been widely used in the past, it
   is a goal of this memo to be as compatible as possible with existing
   practice.  Primarily this means that existing host implementations
   should not interact badly with hosts and routers which implement the
   specifications of this memo, since TOS support is almost non-existent
   in routers which predate this specification.  However, this memo does
   attempt to be compatible with the treatment of IP TOS in OSPF and
   Integrated IS-IS.

   Because the Internet community does not have much experience with
   TOS, it is important that this specification allow easy definition
   and deployment of new and experimental types of service.  This goal
   has had a significant impact on this specification.  In particular,
   it led to the decision to fix permanently the size of the TOS field
   and to the decision that hosts and routers should be able to handle a
   new type of service correctly without having to understand its
   semantics.

   Appendix B of this memo provides a more detailed explanation of the
   rationale behind particular aspects of this specification.

3.  Specification of the Type of Service Octet



   The TOS facility is one of the features of the Type of Service octet
   in the IP datagram header.  The Type of Service octet consists of
   three fields:

                0     1     2     3     4     5     6     7
             +-----+-----+-----+-----+-----+-----+-----+-----+
             |                 |                       |     |
             |   PRECEDENCE    |          TOS          | MBZ |
             |                 |                       |     |
             +-----+-----+-----+-----+-----+-----+-----+-----+

   The first field, labeled "PRECEDENCE" above, is intended to denote
   the importance or priority of the datagram.  This field is not
   discussed in detail in this memo.

   The second field, labeled "TOS" above, denotes how the network should



Almquist                                                        [Page 4]



RFC 1349                    Type of Service                    July 1992


   make tradeoffs between throughput, delay, reliability, and cost.  The
   TOS field is the primary topic of this memo.

   The last field, labeled "MBZ" (for "must be zero") above, is
   currently unused.  The originator of a datagram sets this field to
   zero (unless participating in an Internet protocol experiment which
   makes use of that bit).  Routers and recipients of datagrams ignore
   the value of this field.  This field is copied on fragmentation.

   In the past there has been some confusion about the size of the TOS
   field.  RFC-791 defined it as a three bit field, including bits 3-5
   in the figure above.  It included bit 6 in the MBZ field.  RFC-1122
   added bits 6 and 7 to the TOS field, eliminating the MBZ field.  This
   memo redefines the TOS field to be the four bits shown in the figure
   above.  The reasons for choosing to make the TOS field four bits wide
   can be found in Appendix B.2.

4.  Specification of the TOS Field



   As was stated just above, this memo redefines the TOS field as a four
   bit field.  Also contrary to RFC-791, this memo defines the TOS field
   as a single enumerated value rather than as a set of bits (where each
   bit has its own meaning).  This memo defines the semantics of the
   following TOS field values (expressed as binary numbers):

                    1000   --   minimize delay
                    0100   --   maximize throughput
                    0010   --   maximize reliability
                    0001   --   minimize monetary cost
                    0000   --   normal service

   The values used in the TOS field are referred to in this memo as "TOS
   values", and the value of the TOS field of an IP packet is referred
   to in this memo as the "requested TOS".  The TOS field value 0000 is
   referred to in this memo as the "default TOS."

   Because this specification redefines TOS values to be integers rather
   than sets of bits, computing the logical OR of two TOS values is no
   longer meaningful.  For example, it would be a serious error for a
   router to choose a low delay path for a packet whose requested TOS
   was 1110 simply because the router noted that the former "delay bit"
   was set.

   Although the semantics of values other than the five listed above are
   not defined by this memo, they are perfectly legal TOS values, and
   hosts and routers must not preclude their use in any way.  As will
   become clear after reading the remainder of this memo, only the
   default TOS is in any way special.  A host or router need not (and



Almquist                                                        [Page 5]



RFC 1349                    Type of Service                    July 1992


   except as described in Section 8 should not) make any distinction
   between TOS values whose semantics are defined by this memo and those
   that are not.

   It is important to note the use of the words "minimize" and
   "maximize" in the definitions of values for the TOS field.  For
   example, setting the TOS field to 1000 (minimize delay) does not
   guarantee that the path taken by the datagram will have a delay that
   the user considers "low".  The network will attempt to choose the
   lowest delay path available, based on its (often imperfect)
   information about path delay.  The network will not discard the
   datagram simply because it believes that the delay of the available
   paths is "too high" (actually, the network manager can override this
   behavior through creative use of routing metrics, but this is
   strongly discouraged: setting the TOS field is intended to give
   better service when it is available, rather than to deny service when
   it is not).

5.  Use of the TOS Field in the Internet Protocols



   For the TOS facility to be useful, the TOS fields in IP packets must
   be filled in with reasonable values.  This section discusses how
   protocols above IP choose appropriate values.

   5.1  Internet Control Message Protocol (ICMP)



      ICMP [8,9,12] defines a number of messages for performing error
      reporting and diagnostic functions for the Internet Layer.  This
      section describes how a host or router chooses appropriate TOS
      values for ICMP messages it originates.  The TOS facility also
      affects the origination and processing of ICMP Redirects and ICMP
      Destination Unreachables, but that is the topic of Section 6.

      For purposes of this discussion, it is useful to divide ICMP
      messages into three classes:

       o   ICMP error messages include ICMP message types 3 (Destination
           Unreachable), 4 (Source Quench), 5 (Redirect), 11 (Time
           Exceeded), and 12 (Parameter Problem).

       o   ICMP request messages include ICMP message types 8 (Echo), 10
           (Router Solicitation), 13 (Timestamp), 15 (Information
           Request -- now obsolete), and 17 (Address Mask Request).

       o   ICMP reply messages include ICMP message types 0 (Echo
           Reply), 9 (Router Advertisement), 14 (Timestamp Reply), 16
           (Information Reply -- also obsolete), and 18 (Address Mask
           Reply).



Almquist                                                        [Page 6]



RFC 1349                    Type of Service                    July 1992


      An ICMP error message is always sent with the default TOS (0000).

      An ICMP request message may be sent with any value in the TOS
      field.  A mechanism to allow the user to specify the TOS value to
      be used would be a useful feature in many applications that
      generate ICMP request messages.

      An ICMP reply message is sent with the same value in the TOS field
      as was used in the corresponding ICMP request message.

   5.2  Transport Protocols



      When sending a datagram, a transport protocol uses the TOS
      requested by the application.  There is no requirement that both
      ends of a transport connection use the same TOS.  For example, the
      sending side of a bulk data transfer application should request
      that throughput be maximized, whereas the receiving side might
      request that delay be minimized (assuming that it is primarily
      sending small acknowledgement packets).  It may be useful for a
      transport protocol to provide applications with a mechanism for
      learning the value of the TOS field that accompanied the most
      recently received data.

      It is quite permissible to switch to a different TOS in the middle
      of a connection if the nature of the traffic being generated
      changes.  An example of this would be SMTP, which spends part of
      its time doing bulk data transfer and part of its time exchanging
      short command messages and responses.

      TCP [13] should use the same TOS for datagrams containing only TCP
      control information as it does for datagrams which contain user
      data.  Although it might seem intuitively correct to always
      request that the network minimize delay for segments containing
      acknowledgements but no data, doing so could corrupt TCP's round
      trip time estimates.

   5.3  Application Protocols



      Applications are responsible for choosing appropriate TOS values
      for any traffic they originate.  The Assigned Numbers document
      [15] lists the TOS values to be used by a number of common network
      applications.  For other applications, it is the responsibility of
      the application's designer or programmer to make a suitable
      choice, based on the nature of the traffic to be originated by the
      application.

      It is essential for many sorts of network diagnostic applications,
      and desirable for other applications, that the user of the



Almquist                                                        [Page 7]



RFC 1349                    Type of Service                    July 1992


      application be able to override the TOS value(s) which the
      application would otherwise choose.

      The Assigned Numbers document is revised and reissued
      periodically.  Until RFC-1060, the edition current as this is
      being written, has been superceded, readers should consult
      Appendix A.2 of this memo.

6.  ICMP and the TOS Facility



   Routers communicate routing information to hosts using the ICMP
   protocol [12].  This section describes how support for the TOS
   facility affects the origination and interpretation of ICMP Redirect
   messages and certain types of ICMP Destination Unreachable messages.
   This memo does not define any new extensions to the ICMP protocol.

   6.1  Destination Unreachable



      The ICMP Destination Unreachable message contains a code which
      describes the reason that the destination is unreachable.  There
      are four codes [1,12] which are particularly relevant to the topic
      of this memo:

         0 -- network unreachable
         1 -- host unreachable
        11 -- network unreachable for type of service
        12 -- host unreachable for type of service



      A router generates a code 11 or code 12 Destination Unreachable
      when an unreachable destination (network or host) would have been
      reachable had a different TOS value been specified.  A router
      generates a code 0 or code 1 Destination Unreachable in other
      cases.

      A host receiving a Destination Unreachable message containing any
      of these codes should recognize that it may result from a routing
      transient.  The host should therefore interpret the message as
      only a hint, not proof, that the specified destination is
      unreachable.

      The use of codes 11 and 12 may seem contrary to the statement in
      Section 2 that packets should not be discarded simply because the
      requested TOS cannot be provided.  The rationale for having these
      codes and the limited cases in which they are expected to be used
      are described in Appendix B.5.






Almquist                                                        [Page 8]



RFC 1349                    Type of Service                    July 1992


   6.2  Redirect



      The ICMP Redirect message also includes a code, which specifies
      the class of datagrams to which the Redirect applies.  There are
      currently four codes defined:

         0 -- redirect datagrams for the network
         1 -- redirect datagrams for the host
         2 -- redirect datagrams for the type of service and network
         3 -- redirect datagrams for the type of service and host

      A router generates a code 3 Redirect when the Redirect applies
      only to IP packets which request a particular TOS value.  A router
      generates a code 1 Redirect instead when the the optimal next hop
      on the path to the destination would be the same for any TOS
      value.  In order to minimize the potential for host confusion,
      routers should refrain from using codes 0 and 2 in Redirects
      [3,6].

      Although the current Internet Host specification [1] only requires
      hosts to correctly handle code 0 and code 1 Redirects, a host
      should also correctly handle code 2 and code 3 Redirects, as
      described in Section 7.1 of this memo.  If a host does not, it is
      better for the host to treat code 2 as equivalent to code 0 and
      code 3 as equivalent to code 1 than for the host to simply ignore
      code 2 and code 3 Redirects.

7.  Use of the TOS Field in Routing



   Both hosts and routers should consider the value of the TOS field of
   a datagram when choosing an appropriate path to get the datagram to
   its destination.  The mechanisms for doing so are discussed in this
   section.

   Whether a packet's TOS value actually affects the path it takes
   inside of a particular routing domain is a choice made by the routing
   domain's network manager.  In many routing domains the paths are
   sufficiently homogeneous in nature that there is no reason for
   routers to choose different paths based up the TOS field in a
   datagram.  Inside such a routing domain, the network manager may
   choose to limit the size of the routing database and of routing
   protocol updates by only defining routes for the default (0000) TOS.
   Neither hosts nor routers should need to have any explicit knowledge
   of whether TOS affects routing in the local routing domain.







Almquist                                                        [Page 9]



RFC 1349                    Type of Service                    July 1992


   7.1  Host Routing



      When a host (which is not also a router) wishes to send an IP
      packet to a destination on another network or subnet, it needs to
      choose an appropriate router to send the packet to.  According to
      the IP Architecture, it does so by maintaining a route cache and a
      list of default routers.  Each entry in the route cache lists a
      destination (IP address) and the appropriate router to use to
      reach that destination.  The host learns the information stored in
      its route cache through the ICMP Redirect mechanism.  The host
      learns the list of default routers either from static
      configuration information or by using the ICMP Router Discovery
      mechanism [8].  When the host wishes to send an IP packet, it
      searches its route cache for a route matching the destination
      address in the packet.  If one is found it is used; if not, the
      packet is sent to one of the default routers.  All of this is
      described in greater detail in section 3.3.1 of RFC-1122 [1].

      Adding support for the TOS facility changes the host routing
      procedure only slightly.  In the following, it is assumed that (in
      accordance with the current Internet Host specification [1]) the
      host treats code 0 (redirect datagrams for the network) Redirects
      as if they were code 1 (redirect datagrams for the host)
      Redirects.  Similarly, it is assumed that the host treats code 2
      (redirect datagrams for the network and type of service) Redirects
      as if they were code 3 (redirect datagrams for the host and type
      of service) Redirects.  Readers considering violating these
      assumptions should be aware that long and careful consideration of
      the way in which Redirects are treated is necessary to avoid
      situations where every packet sent to some destination provokes a
      Redirect.  Because these assumptions match the recommendations of
      Internet Host specification, that careful consideration is beyond
      the scope of this memo.

      As was described in Section 6.2, some ICMP Redirects apply only to
      IP packets which request a particular TOS.  Thus, a host (at least
      conceptually) needs to store two types of entries in its route
      cache:

       type 1: { destination, TOS, router }

       type 2: { destination, *, router }

      where type 1 entries result from the receipt of code 3 (or code 1)
      Redirects and type 2 entries result from the receipt of code 2 (or
      code 0) Redirects.





Almquist                                                       [Page 10]



RFC 1349                    Type of Service                    July 1992


      When a host wants to send a packet, it first searches the route
      cache for a type 1 entry whose destination matches the destination
      address of the packet and whose TOS matches the requested TOS in
      the packet.  If it doesn't find one, the host searches its route
      cache again, this time looking for a type 2 entry whose
      destination matches the destination address of the packet.  If
      either of these searches finds a matching entry, the packet is
      sent to the router listed in the matching entry.  Otherwise, the
      packet is sent to one of the routers on the list of default
      routers.

      When a host creates (or updates) a type 2 entry, it must flush
      from its route cache any type 1 entries which have the same
      destination.  This is necessary for correctness, since the type 1
      entry may be obsolete but would continue to be used if it weren't
      flushed because type 1 entries are always preferred over type 2
      entries.

      However, the converse is not true: when a host creates a type 1
      entry, it should not flush a type 2 entry that has the same
      destination.  In this case, the type 1 entry will properly
      override the type 2 entry for packets whose destination address
      and requested TOS match the type 1 entry.  Because the type 2
      entry may well specify the correct router for some TOS values
      other than the one specified in the type 1 entry, saving the type
      2 entry will likely cut down on the number of Redirects which the
      host would otherwise receive.  This savings can potentially be
      substantial if one of the Redirects which was avoided would have
      created a new type 2 entry (thereby causing the new type 1 entry
      to be flushed).  That can happen, for example, if only some of the
      routers on the local net are part of a routing domain that
      computes separate routes for each TOS.

      As an alternative, a host may treat all Redirects as if they were
      code 3 (redirect datagrams for hosts and type of service)
      Redirects.  This alternative allows the host to have only type 1
      route cache entries, thereby simplifying route lookup and
      eliminating the need for the rules in the previous two paragraphs.
      The disadvantage of this approach is that it increases the size of
      the route cache and the amount of Redirect traffic if the host
      sends packets with a variety of requested TOS's to a destination
      for which the host should use the same router regardless of the
      requested TOS.  There is not yet sufficient experience with the
      TOS facility to know whether that disadvantage would be serious
      enough in practice to outweigh the simplicity of this approach.

      Despite RFC-1122, some hosts acquire their routing information by
      "wiretapping" a routing protocol instead of by using the



Almquist                                                       [Page 11]



RFC 1349                    Type of Service                    July 1992


      mechanisms described above.  Such hosts will need to follow the
      procedures described in Section 7.2 (except of course that hosts
      will not send ICMP Destination Unreachables or ICMP Redirects).

   7.2  Forwarding



      A router in the Internet should be able to consider the value of
      the TOS field when choosing an appropriate path over which to
      forward an IP packet.  How a router does this is a part of the
      more general issue of how a router picks appropriate paths.  This
      larger issue can be extremely complex [4], and is beyond the scope
      of this memo.  This discussion should therefore be considered only
      an overview.  Implementors should consult the Router Requirements
      specification [3] and the the specifications of the routing
      protocols they implement for details.

      A router associates a TOS value with each route in its forwarding
      table.  The value can be any of the possible values of the TOS
      field in an IP datagram (including those values whose semantics
      are yet to be defined).  Any routes learned using routing
      protocols which support TOS are assigned appropriate TOS value by
      those protocols.  Routes learned using other routing protocols are
      always assigned the default TOS value (0000).  Static routes have
      their TOS values assigned by the network manager.

      When a router wants to forward a packet, it first looks up the
      destination address in its forwarding table.  This yields a set of
      candidate routes.  The set may be empty (if the destination is
      unreachable), or it may contain one or more routes to the
      destination.  If the set is not empty, the TOS values of the
      routes in the set are examined.  If the set contains a route whose
      TOS exactly matches the TOS field of the packet being forwarded
      then that route is chosen.  If not but the set contains a route
      with the default TOS then that route is chosen.

      If no route is found, or if the the chosen route has an infinite
      metric, the destination is considered to be unreachable.  The
      packet is discarded and an ICMP Destination Unreachable is
      returned to the source.  Normally, the Unreachable uses code 0
      (Network unreachable) or 1 (Host unreachable).  If, however, a
      route to the destination exists which has a different TOS value
      and a non-infinite metric then code 11 (Network unreachable for
      type of service) or code 12 (Host unreachable for type of service)
      must be used instead.







Almquist                                                       [Page 12]



RFC 1349                    Type of Service                    July 1992


8.  Other consequences of TOS



   The TOS field in a datagram primarily affects the path chosen through
   the network, but an implementor may choose to have TOS also affect
   other aspects of how the datagram is handled.  For example, a host or
   router might choose to give preferential queuing on network output
   queues to datagrams which have requested that delay be minimized.
   Similarly, a router forced by overload to discard packets might
   attempt to avoid discarding packets that have requested that
   reliability be maximized.  At least one paper [14] has explored these
   ideas in some detail, but little is known about how well such special
   handling would work in practice.

   Additionally, some Link Layer protocols have their own quality of
   service mechanisms.  When a router or host transmits an IP packet, it
   might request from the Link Layer a quality of service as close as
   possible to the one requested in the TOS field in the IP header.
   Long ago an attempt (RFC-795) was made to codify how this might be
   done, but that document describes Link Layer protocols which have
   since become obsolete and no more recent document on the subject has
   been written.






























Almquist                                                       [Page 13]



RFC 1349                    Type of Service                    July 1992


APPENDIX A.  Updates to Other Specifications



   While this memo is primarily an update to the IP protocol
   specification [11], it also peripherally affects a number of other
   specifications.  This appendix describes those peripheral effects.
   This information is included in an appendix rather than in the main
   body of the document because most if not all of these other
   specifications will be updated in the future.  As that happens, the
   information included in this appendix will become obsolete.

   A.1  RFC-792 (ICMP)



      RFC-792 [12] defines a set of codes indicating reasons why a
      destination is unreachable.  This memo describes the use of two
      additional codes:

        11 -- network unreachable for type of service
        12 -- host unreachable for type of service



      These codes were defined in RFC-1122 [1] but were not included in
      RFC-792.

   A.2  RFC-1060 (Assigned Numbers)



      RFC-1060 [15] describes the old interpretation of the TOS field
      (as three independent bits, with no way to specify that monetary
      cost should be minimized).  Although it is likely obvious how the
      values in RFC-1060 ought to be interpreted in light of this memo,
      the information from that RFC is reproduced here.  The only actual
      changes are for ICMP (to conform to Section 5.1 of this memo) and
      NNTP:

                        ----- Type-of-Service Value -----

         Protocol           TOS Value

         TELNET (1)         1000                 (minimize delay)

         FTP
           Control          1000                 (minimize delay)
           Data (2)         0100                 (maximize throughput)

         TFTP               1000                 (minimize delay)

         SMTP (3)
           Command phase    1000                 (minimize delay)
           DATA phase       0100                 (maximize throughput)




Almquist                                                       [Page 14]



RFC 1349                    Type of Service                    July 1992


                        ----- Type-of-Service Value -----

         Protocol           TOS Value

         Domain Name Service
           UDP Query        1000                 (minimize delay)
           TCP Query        0000
           Zone Transfer    0100                 (maximize throughput)

         NNTP               0001                 (minimize monetary cost)

         ICMP
           Errors           0000
           Requests         0000 (4)
           Responses        <same as request> (4)

         Any IGP            0010                 (maximize reliability)

         EGP                0000

         SNMP               0010                 (maximize reliability)

         BOOTP              0000

         Notes:

          (1) Includes all interactive user protocols (e.g., rlogin).

          (2) Includes all bulk data transfer protocols (e.g., rcp).

          (3) If the implementation does not support changing the TOS
              during the lifetime of the connection, then the
              recommended TOS on opening the connection is the default
              TOS (0000).

          (4) Although ICMP request messages are normally sent with the
              default TOS, there are sometimes good reasons why they
              would be sent with some other TOS value.  An ICMP response
              always uses the same TOS value as was used in the
              corresponding ICMP request message.  See Section 5.1 of
              this memo.

         An application may (at the request of the user) substitute 0001
         (minimize monetary cost) for any of the above values.

         This appendix is expected to be obsoleted by the next revision
         of the Assigned Numbers document.




Almquist                                                       [Page 15]



RFC 1349                    Type of Service                    July 1992


   A.3  RFC-1122 and RFC-1123 (Host Requirements)



      The use of the TOS field by hosts is described in detail in
      RFC-1122 [1] and RFC-1123 [2].  The information provided there is
      still correct, except that:

       (1) The TOS field is four bits wide rather than five bits wide.
           The requirements that refer to the TOS field should refer
           only to the four bits that make up the TOS field.

       (2) An application may set bit 6 of the TOS octet to a non-zero
           value (but still must not set bit 7 to a non-zero value).

      These details will presumably be corrected in the next revision of
      the Host Requirements specification, at which time this appendix
      can be considered obsolete.

   A.4  RFC-1195 (Integrated IS-IS)



      Integrated IS-IS (sometimes known as Dual IS-IS) has multiple
      metrics for each route.  Which of the metrics is used to route a
      particular IP packet is determined by the TOS field in the packet.
      This is described in detail in section 3.5 of RFC-1195 [7].

      The mapping from the value of the TOS field to an appropriate
      Integrated IS-IS metric is described by a table in that section.
      Although the specification in this memo is intended to be
      substantially compatible with Integrated IS-IS, the extension of
      the TOS field to four bits and the addition of a TOS value
      requesting "minimize monetary cost" require minor modifications to
      that table, as shown here:

         The IP TOS octet is mapped onto the four available metrics as
         follows:

         Bits 0-2 (Precedence): (unchanged from RFC-1195)

         Bits 3-6 (TOS):

            0000    (all normal)               Use default metric
            1000    (minimize delay)           Use delay metric
            0100    (maximize throughput)      Use default metric
            0010    (maximize reliability)     Use reliability metric
            0001    (minimize monetary cost)   Use cost metric
            other                              Use default metric

         Bit 7 (MBZ): This bit is ignored by Integrated IS-IS.




Almquist                                                       [Page 16]



RFC 1349                    Type of Service                    July 1992


      It is expected that the next revision of the Integrated IS-IS
      specification will include this corrected table, at which time
      this appendix can be considered obsolete.

   A.5  RFC-1247 (OSPF) and RFC-1248 (OSPF MIB)



      Although the specification in this memo is intended to be
      substantially compatible with OSPF, the extension of the TOS field
      to four bits requires minor modifications to the section that
      describes the encoding of TOS values in Link State Advertisements,
      described in section 12.3 of RFC-1247 [10].  The encoding is
      summarized in Table 17 of that memo; what follows is an updated
      version of table 17.  The numbers in the first column are decimal
      integers, and the numbers in the second column are binary TOS
      values:

                OSPF encoding   TOS
                _____________________________________________

                0               0000   normal service
                2               0001   minimize monetary cost
                4               0010   maximize reliability
                6               0011
                8               0100   maximize throughput
                10              0101
                12              0110
                14              0111
                16              1000   minimize delay
                18              1001
                20              1010
                22              1011
                24              1100
                26              1101
                28              1110
                30              1111

      The OSPF MIB, described in RFC-1248 [5], is entirely consistent
      with this memo except for the textual comment which describes the
      mapping of the old TOS flag bits into TOSType values.  TOSType
      values use the same encoding of TOS values as OSPF's Link State
      Advertisements do, so the above table also describes the mapping
      between TOSType values (the first column) and TOS field values
      (the second column).

      If RFC-1247 and RFC-1248 are revised in the future, it is expected
      that this information will be incorporated into the revised
      versions.  At that time, this appendix may be considered obsolete.




Almquist                                                       [Page 17]



RFC 1349                    Type of Service                    July 1992


APPENDIX B.  Rationale



   The main body of this memo has described the details of how TOS
   facility works.  This appendix is for those who wonder why it works
   that way.

   Much of what is in this document can be explained by the simple fact
   that the goal of this document is to provide a clear and complete
   specification of the existing TOS facility rather than to design from
   scratch a new quality of service mechanism for IP.  While this memo
   does amend the facility in some small and carefully considered ways
   discussed below, the desirability of compatibility with existing
   specifications and uses of the TOS facility [1,2,7,10,11] was never
   in doubt.  This goal of backwards compatibility determined the broad
   outlines and many of the details of this specification.

   Much of the rest of this specification was determined by two
   additional goals, which were described more fully in Section 2.  The
   first was that hosts should never be penalized for using the TOS
   facility, since that would likely ensure that it would never be
   widely deployed.  The second was that the specification should make
   it easy, or at least possible, to define and deploy new types of
   service in the future.

   The three goals above did not eliminate all need for engineering
   choices, however, and in a few cases the goals proved to be in
   conflict with each other.  The remainder of this appendix discusses
   the rationale behind some of these engineering choices.

   B.1  The Minimize Monetary Cost TOS Value



      Because the Internet is becoming increasingly commercialized, a
      number of participants in the IETF's Router Requirements Working
      Group felt it would be important to have a TOS value which would
      allow a user to declare that monetary cost was more important than
      other qualities of the service.

      There was considerable debate over what exactly this value should
      mean.  Some felt, for example, that the TOS value should mean
      "must not cost money".  This was rejected for several reasons.
      Because it would request a particular level of service (cost = 0)
      rather than merely requesting that some service attribute be
      minimized or maximized, it would not only philosophically at odds
      with the other TOS values but would require special code in both
      hosts and routers.  Also, it would not be helpful to users who
      want their packets to travel via the least-cost path but can
      accept some level of cost when necessary.  Finally, since whether
      any particular routing domain considers the TOS field when routing



Almquist                                                       [Page 18]



RFC 1349                    Type of Service                    July 1992


      is a choice made by the network manager, a user requiring a free
      path might not get one if the packet has to pass through a routing
      domain that does not consider TOS in its routing decisions.

      Some proposed a slight variant: a TOS value which would mean "I am
      willing to pay money to have this packet delivered".  This
      proposal suffers most of the same shortcomings as the previous one
      and turns out to have an additional interesting quirk: because of
      the algorithms specified in Section 7.2, any packet which used
      this TOS value would prefer links that cost money over equally
      good free links.  Thus, such a TOS value would almost be
      equivalent to a "maximize monetary cost" value!

      It seems likely that in the future users may need some mechanism
      to express the maximum amount they are willing to pay to have a
      packet delivered.  However, an IP option would be a more
      appropriate mechanism, since there are precedents for having IP
      options that all routers are required to honor, and an IP option
      could include parameters such as the maximum amount the user was
      willing to pay.  Thus, the TOS value defined in this memo merely
      requests that the network "minimize monetary cost".

   B.2  The Specification of the TOS Field



      There were four goals that guided the decision to have a four bit
      TOS field and the specification of that field's values:

       (1) To define a new type of service requesting that the network
           "minimize monetary cost"

       (2) To remain as compatible as possible with existing
           specifications and uses of the TOS facility

       (3) To allow for the definition and deployment of new types of
           service in the future

       (4) To permanently fix the size of the TOS field

      The last goal may seem surprising, but turns out to be necessary
      for routing to work correctly when new types of service are
      deployed.  If routers have different ideas about the size of the
      TOS field they make inconsistent decisions that may lead to
      routing loops.

      At first glance goals (3) and (4) seem to be pretty much mutually
      exclusive.  The IP header currently has only three unused bits, so
      at most three new type of service bits could be defined without
      resorting to the impractical step of changing the IP header



Almquist                                                       [Page 19]



RFC 1349                    Type of Service                    July 1992


      format.  Since one of them would need to be allocated to meet goal
      (1), at most two bits could be reserved for new or experimental
      types of service.  Not only is it questionable whether two would
      be enough, but it is improbable that the IETF and IAB would allow
      all of the currently unused bits to be permanently reserved for
      types of service which might or might or might not ever be
      defined.

      However, some (if not most of) the possible combinations of the
      individual bits would not be useful.  Clearly, setting all of the
      bits would be equivalent to setting none of the bits, since
      setting all of the bits would indicate that none of the types of
      optimization was any more important than any of the others.
      Although one could perhaps assign reasonable semantics to most
      pairs of bits, it is unclear that the range of network service
      provided by various paths could usefully be subdivided in so fine
      a manner.  If some of these non-useful combinations of bits could
      be assigned to new types of service then it would be possible to
      meet goal (3) and goal (4) without having to use up all of the
      remaining reserved bits in the IP header.  The obvious way to do
      that was to change the interpretation of TOS values so that they
      were integers rather than independently settable bits.

      The integers were chosen to be compatible with the bit definitions
      found in RFC-791.  Thus, for example, setting the TOS field to
      1000 (minimize delay) sets bit 3 of the Type of Service octet; bit
      3 is defined as the Low Delay bit in RFC-791.  This memo only
      defines values which correspond to setting a single one of the
      RFC-791 bits, since setting multiple TOS bits does not seem to be
      a common practice.  According to [15], none of the common TCP/IP
      applications currently set multiple TOS bits.  However, TOS values
      corresponding to particular combinations of the RFC-791 bits could
      be defined if and when they are determined to be useful.

      The new TOS value for "minimize monetary cost" needed to be one
      which would not be too terribly misconstrued by preexisting
      implementations.  This seemed to imply that the value should be
      one which left all of the RFC-791 bits clear.  That would require
      expanding the TOS field, but would allow old implementations to
      treat packets which request minimization of monetary cost (TOS
      0001) as if they had requested the default TOS.  This is not a
      perfect solution since (as described above) changing the size of
      the TOS field could cause routing loops if some routers were to
      route based on a three bit TOS field and others were to route
      based on a four bit TOS field.  Fortunately, this should not be
      much of a problem in practice because routers which route based on
      a three bit TOS field are very rare as this is being written and
      will only become more so once this specification is published.



Almquist                                                       [Page 20]



RFC 1349                    Type of Service                    July 1992


      Because of those considerations, and also in order to allow a
      reasonable number of TOS values for future definition, it seemed
      desirable to expand the TOS field.  That left the question of how
      much to expand it.  Expanding it to five bits would allow
      considerable future expansion (27 new TOS values) and would be
      consistent with Host Requirements, but would reduce to one the
      number of reserved bits in the IP header.  Expanding the TOS field
      to four bits would restrict future expansion to more modest levels
      (11 new TOS values), but would leave an additional IP header bit
      free.  The IETF's Router Requirements Working Group concluded that
      a four bits wide TOS field allow enough values for future use and
      that consistency with Host Requirements was inadequate
      justification for unnecessarily increasing the size of the TOS
      field.

   B.3  The Choice of Weak TOS Routing



      "Ruminations on the Next Hop" [4] describes three alternative ways
      of routing based on the TOS field.  Briefly, they are:

       (1) Strong TOS --
           a route may be used only if its TOS exactly matches the TOS
           in the datagram being routed.  If there is no route with the
           requested TOS, the packet is discarded.

       (2) Weak TOS --
           like Strong TOS, except that a route with the default TOS
           (0000) is used if there is no route that has the requested
           TOS.  If there is no route with either the requested TOS or
           the default TOS, the packet is discarded.

       (3) Very Weak TOS --
           like Weak TOS, except that a route with the numerically
           smallest TOS is used if there is no route that has either the
           requested TOS or the default TOS.

      This specification has adopted Weak TOS.

      Strong TOS was quickly rejected.  Because it requires that each
      router a packet traverses have a route with the requested TOS,
      packets which requested non-zero TOS values would have (at least
      until the TOS facility becomes widely used) a high probability of
      being discarded as undeliverable.  This violates the principle
      (described in Section 2) that hosts should not be penalized for
      choosing non-zero TOS values.

      The choice between Weak TOS and Very Weak TOS was not as
      straightforward.  Weak TOS was chosen because it is slightly



Almquist                                                       [Page 21]



RFC 1349                    Type of Service                    July 1992


      simpler to implement and because it is consistent with the OSPF
      and Integrated IS-IS specifications.  In addition, many dislike
      Very Weak TOS because its algorithm for choosing a route when none
      of the available routes have either the requested or the default
      TOS cannot be justified by intuition (there is no reason to
      believe that having a numerically smaller TOS makes a route
      better).  Since a router would need to understand the semantics of
      all of the TOS values to make a more intelligent choice, there
      seems to be no reasonable way to fix this particular deficiency of
      Very Weak TOS.

      In practice it is expected that the choice between Weak TOS and
      Very Weak TOS will make little practical difference, since (except
      where the network manager has intentionally set things up
      otherwise) there will be a route with the default TOS to any
      destination for which there is a route with any other TOS.

   B.4  The Retention of Longest Match Routing



      An interesting issue is how early in the route choice process TOS
      should be considered.  There seem to be two obvious possibilities:

       (1) Find the set of routes that best match the destination
           address of the packet.  From among those, choose the route
           which best matches the requested TOS.

       (2) Find the set of routes that best match the requested TOS.
           From among those, choose the route which best matches the
           destination address of the packet.

      The two approaches are believed to support an identical set of
      routing policies.  Which of the two allows the simpler
      configuration and minimizes the amount of routing information that
      needs to be passed around seems to depend on the topology, though
      some believe that the second option has a slight edge in this
      regard.

      Under the first option, if the network manager neglects some
      pieces of the configuration the likely consequence is that some
      packets which would benefit from TOS-specific routes will be
      routed as if they had requested the default TOS.  Under the second
      option, however, a network manager can easily (accidently)
      configure things in such a way that packets which request a
      certain TOS and should be delivered locally will instead follow a
      default route for that TOS and be dumped into the Internet.  Thus,
      the first option would seem to have a slight edge with regard to
      robustness in the face of errors by the network manager.




Almquist                                                       [Page 22]



RFC 1349                    Type of Service                    July 1992


      It has been also been suggested that the first option provides the
      additional benefit of allowing loop-free routing in routing
      domains which contain both routers that consider TOS in their
      routing decisions and routers that do not.  Whether that is true
      in all cases is unknown.  It is certainly the case, however, that
      under the second option it would not work to mix routers that
      consider TOS and routers which do not in the same routing domain.

      All in all, there were no truly compelling arguments for choosing
      one way or the other, but it was nontheless necessary to make a
      choice: if different routers were to make the choice differently,
      chaos (in the form of routing loops) would result.  The mechanisms
      specified in this memo reflect the first option because that will
      probably be more intuitive to most network managers.  Internet
      routing has traditionally chosen the route which best matches the
      destination address, with other mechanisms serving merely as tie-
      breakers.  The first option is consistent with that tradition.

   B.5  The Use of Destination Unreachable



      Perhaps the most contentious and least defensible part of this
      specification is that a packet can be discarded because the
      destination is considered to be unreachable even though a packet
      to the same destination but requesting a different TOS would have
      been deliverable.  This would seem to fall perilously close to
      violating the principle that hosts should never be penalized for
      requesting non-default TOS values in packets they originate.

      This can happen in only three, somewhat unusual, cases:

       (1) There is a route to the packet's destination which has the
           TOS value requested in the packet, but the route has an
           infinite metric.

       (2) The only routes to the packet's destination have TOS values
           other than the one requested in the packet.  One of them has
           the default TOS, but it has an infinite metric.

       (3) The only routes to the packet's destination have TOS values
           other than the one requested in the packet.  None of them
           have the default TOS.

      It is commonly accepted that a router which has a default route
      should nonetheless discard a packet if the router has a more
      specific route to the destination in its forwarding table but that
      route has an infinite metric.  The first two cases seem to be
      analogous to that rule.




Almquist                                                       [Page 23]



RFC 1349                    Type of Service                    July 1992


      In addition, it is worth noting that, except perhaps during brief
      transients resulting from topology changes, routes with infinite
      metrics occur only as the result of deliberate action (or serious
      error) on the part of the network manager.  Thus, packets are
      unlikely to be discarded unless the network manager has taken
      deliberate action to cause them to be.  Some people believe that
      this is an important feature of the specification, allowing the
      network to (for example) keep packets which have requested that
      cost be minimized off of a link that is so expensive that the
      network manager feels confident that the users would want their
      packets to be dropped.  Others (including the author of this memo)
      believe that this "feature" will prove not to be useful, and that
      other mechanisms may be required for access controls on links, but
      couldn't justify changing this specification in the ways necessary
      to eliminate the "feature".

      Case (3) above is more problematic.  It could have been avoided by
      using Very Weak TOS, but that idea was rejected for the reasons
      discussed in Appendix B.3.  Some suggested that case (3) could be
      fixed by relaxing longest match routing (described in Appendix
      B.4), but that idea was rejected because it would add complexity
      to routers without necessarily making their routing choices
      particularly more intuitive.  It is also worth noting that this is
      another case that a network manager has to try rather hard to
      create: since OSPF and Integrated IS-IS both enforce the
      constraint that there must be a route with the default TOS to any
      destination for which there is a route with a non-zero TOS, a
      network manager would have to await the development of a new
      routing protocol or create the problem with static routes.  The
      eventual conclusion was that any fix to case (3) was worse than
      the problem.

APPENDIX C.  Limitations of the TOS Mechanism



   It is important to note that the TOS facility has some limitations.
   Some are consequences of engineering choices made in this
   specification.  Others, referred to as "inherent limitations" below,
   could probably not have been avoided without either replacing the TOS
   facility defined in RFC-791 or accepting that things wouldn't work
   right until all routers in the Internet supported the TOS facility.

   C.1  Inherent Limitations



      The most important of the inherent limitations is that the TOS
      facility is strictly an advisory mechanism.  It is not an
      appropriate mechanism for requesting service guarantees.  There
      are two reasons why this is so:




Almquist                                                       [Page 24]



RFC 1349                    Type of Service                    July 1992


       (1) Not all networks will consider the value of the TOS field
           when deciding how to handle and route packets.  Partly this
           is a transition issue: there will be a (probably lengthy)
           period when some networks will use equipment that predates
           this specification.  Even long term, however, many networks
           will not be able to provide better service by considering the
           value of the TOS field.  For example, the best path through a
           network composed of a homogeneous collection of
           interconnected LANs is probably the same for any possible TOS
           value.  Inside such a network, it would make little sense to
           require routers and routing protocols to do the extra work
           needed to consider the value of the TOS field when forwarding
           packets.

       (2) The TOS mechanism is not powerful enough to allow an
           application to quantify the level of service it desires.  For
           example, an application may use the TOS field to request that
           the network choose a path which maximizes throughput, but
           cannot use that mechanism to say that it needs or wants a
           particular number of kilobytes or megabytes per second.
           Because the network cannot know what the application
           requires, it would be inappropriate for the network to decide
           to discard a packet which requested maximal throughput
           because no "high throughput" path was available.

      The inability to provide resource guarantees is a serious drawback
      for certain kinds of network applications.  For example, a system
      using packetized voice simply creates network congestion when the
      available bandwidth is inadequate to deliver intelligible speech.
      Likewise, the network oughtn't even bother to deliver a voice
      packet that has suffered more delay in the network than the
      application can tolerate.  Unfortunately, resource guarantees are
      problematic in connectionless networks.  Internet researchers are
      actively studying this problem, and are optimistic that they will
      be able to invent ways in which the Internet Architecture can
      evolve to support resource guarantees while preserving the
      advantages of connectionless networking.

   C.2  Limitations of this Specification



      There are a couple of additional limitations of the TOS facility
      which are not inherent limitations but instead are consequences of
      engineering choices made in this specification:

       (1) Routing is not really optimal for some TOS values.  This is
           because optimal routing for those TOS values would require
           that routing protocols be cognizant of the semantics of the
           TOS values and use special algorithms to compute routes for



Almquist                                                       [Page 25]



RFC 1349                    Type of Service                    July 1992


           them.  For example, routing protocols traditionally compute
           the metric for a path by summing the costs of the individual
           links that make up the path.  However, to maximize
           reliability, a routing protocol would instead have to compute
           a metric which was the product of the probabilities of
           successful delivery over each of the individual links in the
           path.  While this limitation is in some sense a limitation of
           current routing protocols rather than of this specification,
           this specification contributes to the problem by specifying
           that there are a number of legal TOS values that have no
           currently defined semantics.

       (2) This specification assumes that network managers will do "the
           right thing".  If a routing domain uses TOS, the network
           manager must configure the routers in such a way that a
           reasonable path is chosen for each TOS.  While this ought not
           to be terribly difficult, a network manager could accidently
           or intentionally violate our rule that using the TOS facility
           should provide service at least as good as not using it.
































Almquist                                                       [Page 26]



RFC 1349                    Type of Service                    July 1992


References

  [1]   Internet Engineering Task Force (R. Braden, Editor),
        "Requirements for Internet Hosts -- Communication Layers", RFC
        1122, USC/Information Sciences Institute, October 1989.

  [2]   Internet Engineering Task Force (R. Braden, Editor),
        "Requirements for Internet Hosts -- Application and Support",
        RFC 1123, USC/Information Sciences Institute, October 1989.

  [3]   Almquist, P., "Requirements for IP Routers", Work in progress.

  [4]   Almquist, P., "Ruminations on the Next Hop", Work in progress.

  [5]   Baker, F. and R. Coltun, "OSPF Version 2 Management Information
        Base", RFC 1248, ACC, Computer Science Center, August 1991.

  [6]   Braden, R. and J. Postel, "Requirements for Internet Gateways",
        RFC 1009, USC/Information Sciences Institute, June 1987.

  [7]   Callon, R., "Use of OSI IS-IS for Routing in TCP/IP and Dual
        Environments", RFC 1195, Digital Equipment Corporation, December
        1990.

  [8]   Deering, S., "ICMP Router Discovery Messages", RFC 1256, Xerox
        PARC, September 1991.

  [9]   Mogul, J. and J. Postel, "Internet Standard Subnetting
        Procedure", RFC 950, USC/Information Sciences Institute, August
        1985.

 [10]   Moy, J., "OSPF Version 2", RFC 1247, Proteon, Inc., July 1991.

 [11]   Postel, J., "Internet Protocol", RFC 791, DARPA, September 1981.

 [12]   Postel, J., "Internet Control Message Protocol", RFC 792, DARPA,
        September 1981.

 [13]   Postel, J., "Transmission Control Protocol", RFC 793, DARPA,
        September 1981.

 [14]   Prue, W. and J. Postel, "A Queuing Algorithm to Provide Type-
        of-Service for IP Links", RFC 1046, USC/Information Sciences
        Institute, February 1988.

 [15]   Reynolds, J. and J. Postel, "Assigned Numbers", RFC 1060,
        USC/Information Sciences Institute, March 1990.




Almquist                                                       [Page 27]



RFC 1349                    Type of Service                    July 1992


Acknowledgements



   Some of the ideas presented in this memo are based on discussions
   held by the IETF's Router Requirements Working Group.  Much of the
   specification of the treatment of Type of Service by hosts is merely
   a restatement of the ideas of the IETF's former Host Requirements
   Working Group, as captured in RFC-1122 and RFC-1123.  The author is
   indebted to John Moy and Ross Callon for their assistance and
   cooperation in achieving consistency among the OSPF specification,
   the Integrated IS-IS specification, and this memo.

   This memo has been substantially improved as the result of thoughtful
   comments from a number of reviewers, including Dave Borman, Bob
   Braden, Ross Callon, Vint Cerf, Noel Chiappa, Deborah Estrin, Phill
   Gross, Bob Hinden, Steve Huston, Jon Postel, Greg Vaudreuil, John
   Wobus, and the Router Requirements Working Group.

   The initial work on this memo was done while its author was an
   employee of BARRNet.  Their support is gratefully acknowledged.

Security Considerations

   This memo does not explicitly discuss security issues.  The author
   does not believe that the specifications in this memo either weaken
   or enhance the security of the IP Protocol or of the other protocols
   mentioned herein.

Author's Address



   Philip Almquist
   214 Cole Street, Suite 2
   San Francisco, CA 94117-1916

   Phone: 415-752-2427

   Email: almquist@Jessica.Stanford.EDU















Almquist                                                       [Page 28]