RFC 1981
This document is obsolete. Please refer to RFC 8201.






Network Working Group                                          J. McCann
Request for Comments: 1981                 Digital Equipment Corporation
Category: Standards Track                                     S. Deering
                                                              Xerox PARC
                                                                J. Mogul
                                           Digital Equipment Corporation
                                                             August 1996


                  Path MTU Discovery for IP version 6

Status of this Memo



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

Abstract



   This document describes Path MTU Discovery for IP version 6.  It is
   largely derived from RFC 1191, which describes Path MTU Discovery for
   IP version 4.

Table of Contents



   1. Introduction.................................................2
   2. Terminology..................................................2
   3. Protocol overview............................................3
   4. Protocol Requirements........................................4
   5. Implementation Issues........................................5
   5.1. Layering...................................................5
   5.2. Storing PMTU information...................................6
   5.3. Purging stale PMTU information.............................8
   5.4. TCP layer actions..........................................9
   5.5. Issues for other transport protocols......................11
   5.6. Management interface......................................12
   6. Security Considerations.....................................12
   Acknowledgements...............................................13
   Appendix A - Comparison to RFC 1191............................14
   References.....................................................14
   Authors' Addresses.............................................15








McCann, Deering & Mogul     Standards Track                     [Page 1]

RFC 1981              Path MTU Discovery for IPv6            August 1996


1. Introduction



   When one IPv6 node has a large amount of data to send to another
   node, the data is transmitted in a series of IPv6 packets.  It is
   usually preferable that these packets be of the largest size that can
   successfully traverse the path from the source node to the
   destination node.  This packet size is referred to as the Path MTU
   (PMTU), and it is equal to the minimum link MTU of all the links in a
   path.  IPv6 defines a standard mechanism for a node to discover the
   PMTU of an arbitrary path.

   IPv6 nodes SHOULD implement Path MTU Discovery in order to discover
   and take advantage of paths with PMTU greater than the IPv6 minimum
   link MTU [IPv6-SPEC].  A minimal IPv6 implementation (e.g., in a boot
   ROM) may choose to omit implementation of Path MTU Discovery.

   Nodes not implementing Path MTU Discovery use the IPv6 minimum link
   MTU defined in [IPv6-SPEC] as the maximum packet size.  In most
   cases, this will result in the use of smaller packets than necessary,
   because most paths have a PMTU greater than the IPv6 minimum link
   MTU.  A node sending packets much smaller than the Path MTU allows is
   wasting network resources and probably getting suboptimal throughput.

2. Terminology



   node        - a device that implements IPv6.

   router      - a node that forwards IPv6 packets not explicitly
                 addressed to itself.

   host        - any node that is not a router.

   upper layer - a protocol layer immediately above IPv6.  Examples are
                 transport protocols such as TCP and UDP, control
                 protocols such as ICMP, routing protocols such as OSPF,
                 and internet or lower-layer protocols being "tunneled"
                 over (i.e., encapsulated in) IPv6 such as IPX,
                 AppleTalk, or IPv6 itself.

   link        - a communication facility or medium over which nodes can
                 communicate at the link layer, i.e., the layer
                 immediately below IPv6.  Examples are Ethernets (simple
                 or bridged); PPP links; X.25, Frame Relay, or ATM
                 networks; and internet (or higher) layer "tunnels",
                 such as tunnels over IPv4 or IPv6 itself.

   interface   - a node's attachment to a link.




McCann, Deering & Mogul     Standards Track                     [Page 2]

RFC 1981              Path MTU Discovery for IPv6            August 1996


   address     - an IPv6-layer identifier for an interface or a set of
                 interfaces.

   packet      - an IPv6 header plus payload.

   link MTU    - the maximum transmission unit, i.e., maximum packet
                 size in octets, that can be conveyed in one piece over
                 a link.

   path        - the set of links traversed by a packet between a source
                 node and a destination node

   path MTU    - the minimum link MTU of all the links in a path between
                 a source node and a destination node.

   PMTU        - path MTU

   Path MTU
   Discovery   - process by which a node learns the PMTU of a path

   flow        - a sequence of packets sent from a particular source
                 to a particular (unicast or multicast) destination for
                 which the source desires special handling by the
                 intervening routers.

   flow id     - a combination of a source address and a non-zero
                 flow label.

3. Protocol overview



   This memo describes a technique to dynamically discover the PMTU of a
   path.  The basic idea is that a source node initially assumes that
   the PMTU of a path is the (known) MTU of the first hop in the path.
   If any of the packets sent on that path are too large to be forwarded
   by some node along the path, that node will discard them and return
   ICMPv6 Packet Too Big messages [ICMPv6].  Upon receipt of such a
   message, the source node reduces its assumed PMTU for the path based
   on the MTU of the constricting hop as reported in the Packet Too Big
   message.

   The Path MTU Discovery process ends when the node's estimate of the
   PMTU is less than or equal to the actual PMTU.  Note that several
   iterations of the packet-sent/Packet-Too-Big-message-received cycle
   may occur before the Path MTU Discovery process ends, as there may be
   links with smaller MTUs further along the path.

   Alternatively, the node may elect to end the discovery process by
   ceasing to send packets larger than the IPv6 minimum link MTU.



McCann, Deering & Mogul     Standards Track                     [Page 3]

RFC 1981              Path MTU Discovery for IPv6            August 1996


   The PMTU of a path may change over time, due to changes in the
   routing topology.  Reductions of the PMTU are detected by Packet Too
   Big messages.  To detect increases in a path's PMTU, a node
   periodically increases its assumed PMTU.  This will almost always
   result in packets being discarded and Packet Too Big messages being
   generated, because in most cases the PMTU of the path will not have
   changed.  Therefore, attempts to detect increases in a path's PMTU
   should be done infrequently.

   Path MTU Discovery supports multicast as well as unicast
   destinations.  In the case of a multicast destination, copies of a
   packet may traverse many different paths to many different nodes.
   Each path may have a different PMTU, and a single multicast packet
   may result in multiple Packet Too Big messages, each reporting a
   different next-hop MTU.  The minimum PMTU value across the set of
   paths in use determines the size of subsequent packets sent to the
   multicast destination.

   Note that Path MTU Discovery must be performed even in cases where a
   node "thinks" a destination is attached to the same link as itself.
   In a situation such as when a neighboring router acts as proxy [ND]
   for some destination, the destination can to appear to be directly
   connected but is in fact more than one hop away.

4. Protocol Requirements



   As discussed in section 1, IPv6 nodes are not required to implement
   Path MTU Discovery.  The requirements in this section apply only to
   those implementations that include Path MTU Discovery.

   When a node receives a Packet Too Big message, it MUST reduce its
   estimate of the PMTU for the relevant path, based on the value of the
   MTU field in the message.  The precise behavior of a node in this
   circumstance is not specified, since different applications may have
   different requirements, and since different implementation
   architectures may favor different strategies.

   After receiving a Packet Too Big message, a node MUST attempt to
   avoid eliciting more such messages in the near future.  The node MUST
   reduce the size of the packets it is sending along the path.  Using a
   PMTU estimate larger than the IPv6 minimum link MTU may continue to
   elicit Packet Too Big messages.  Since each of these messages (and
   the dropped packets they respond to) consume network resources, the
   node MUST force the Path MTU Discovery process to end.

   Nodes using Path MTU Discovery MUST detect decreases in PMTU as fast
   as possible.  Nodes MAY detect increases in PMTU, but because doing
   so requires sending packets larger than the current estimated PMTU,



McCann, Deering & Mogul     Standards Track                     [Page 4]

RFC 1981              Path MTU Discovery for IPv6            August 1996


   and because the likelihood is that the PMTU will not have increased,
   this MUST be done at infrequent intervals.  An attempt to detect an
   increase (by sending a packet larger than the current estimate) MUST
   NOT
be done less than 5 minutes after a Packet Too Big message has
   been received for the given path.  The recommended setting for this
   timer is twice its minimum value (10 minutes).

   A node MUST NOT reduce its estimate of the Path MTU below the IPv6
   minimum link MTU.

      Note: A node may receive a Packet Too Big message reporting a
      next-hop MTU that is less than the IPv6 minimum link MTU.  In that
      case, the node is not required to reduce the size of subsequent
      packets sent on the path to less than the IPv6 minimun link MTU,
      but rather must include a Fragment header in those packets [IPv6-
      SPEC].

   A node MUST NOT increase its estimate of the Path MTU in response to
   the contents of a Packet Too Big message.  A message purporting to
   announce an increase in the Path MTU might be a stale packet that has
   been floating around in the network, a false packet injected as part
   of a denial-of-service attack, or the result of having multiple paths
   to the destination, each with a different PMTU.

5. Implementation Issues



   This section discusses a number of issues related to the
   implementation of Path MTU Discovery.  This is not a specification,
   but rather a set of notes provided as an aid for implementors.

   The issues include:

   - What layer or layers implement Path MTU Discovery?

   - How is the PMTU information cached?

   - How is stale PMTU information removed?

   - What must transport and higher layers do?

5.1. Layering



   In the IP architecture, the choice of what size packet to send is
   made by a protocol at a layer above IP.  This memo refers to such a
   protocol as a "packetization protocol".  Packetization protocols are
   usually transport protocols (for example, TCP) but can also be
   higher-layer protocols (for example, protocols built on top of UDP).




McCann, Deering & Mogul     Standards Track                     [Page 5]

RFC 1981              Path MTU Discovery for IPv6            August 1996


   Implementing Path MTU Discovery in the packetization layers
   simplifies some of the inter-layer issues, but has several drawbacks:
   the implementation may have to be redone for each packetization
   protocol, it becomes hard to share PMTU information between different
   packetization layers, and the connection-oriented state maintained by
   some packetization layers may not easily extend to save PMTU
   information for long periods.

   It is therefore suggested that the IP layer store PMTU information
   and that the ICMP layer process received Packet Too Big messages.
   The packetization layers may respond to changes in the PMTU, by
   changing the size of the messages they send.  To support this
   layering, packetization layers require a way to learn of changes in
   the value of MMS_S, the "maximum send transport-message size".  The
   MMS_S is derived from the Path MTU by subtracting the size of the
   IPv6 header plus space reserved by the IP layer for additional
   headers (if any).

   It is possible that a packetization layer, perhaps a UDP application
   outside the kernel, is unable to change the size of messages it
   sends.  This may result in a packet size that exceeds the Path MTU.
   To accommodate such situations, IPv6 defines a mechanism that allows
   large payloads to be divided into fragments, with each fragment sent
   in a separate packet (see [IPv6-SPEC] section "Fragment Header").
   However, packetization layers are encouraged to avoid sending
   messages that will require fragmentation (for the case against
   fragmentation, see [FRAG]).

5.2. Storing PMTU information



   Ideally, a PMTU value should be associated with a specific path
   traversed by packets exchanged between the source and destination
   nodes.  However, in most cases a node will not have enough
   information to completely and accurately identify such a path.
   Rather, a node must associate a PMTU value with some local
   representation of a path.  It is left to the implementation to select
   the local representation of a path.

   In the case of a multicast destination address, copies of a packet
   may traverse many different paths to reach many different nodes.  The
   local representation of the "path" to a multicast destination must in
   fact represent a potentially large set of paths.

   Minimally, an implementation could maintain a single PMTU value to be
   used for all packets originated from the node.  This PMTU value would
   be the minimum PMTU learned across the set of all paths in use by the
   node.  This approach is likely to result in the use of smaller
   packets than is necessary for many paths.



McCann, Deering & Mogul     Standards Track                     [Page 6]

RFC 1981              Path MTU Discovery for IPv6            August 1996


   An implementation could use the destination address as the local
   representation of a path.  The PMTU value associated with a
   destination would be the minimum PMTU learned across the set of all
   paths in use to that destination.  The set of paths in use to a
   particular destination is expected to be small, in many cases
   consisting of a single path.  This approach will result in the use of
   optimally sized packets on a per-destination basis.  This approach
   integrates nicely with the conceptual model of a host as described in
   [ND]: a PMTU value could be stored with the corresponding entry in
   the destination cache.

   If flows [IPv6-SPEC] are in use, an implementation could use the flow
   id as the local representation of a path.  Packets sent to a
   particular destination but belonging to different flows may use
   different paths, with the choice of path depending on the flow id.
   This approach will result in the use of optimally sized packets on a
   per-flow basis, providing finer granularity than PMTU values
   maintained on a per-destination basis.

   For source routed packets (i.e. packets containing an IPv6 Routing
   header [IPv6-SPEC]), the source route may further qualify the local
   representation of a path.  In particular, a packet containing a type
   0 Routing header in which all bits in the Strict/Loose Bit Map are
   equal to 1 contains a complete path specification.  An implementation
   could use source route information in the local representation of a
   path.

      Note: Some paths may be further distinguished by different
      security classifications.  The details of such classifications are
      beyond the scope of this memo.

   Initially, the PMTU value for a path is assumed to be the (known) MTU
   of the first-hop link.

   When a Packet Too Big message is received, the node determines which
   path the message applies to based on the contents of the Packet Too
   Big message.  For example, if the destination address is used as the
   local representation of a path, the destination address from the
   original packet would be used to determine which path the message
   applies to.

      Note: if the original packet contained a Routing header, the
      Routing header should be used to determine the location of the
      destination address within the original packet.  If Segments Left
      is equal to zero, the destination address is in the Destination
      Address field in the IPv6 header.  If Segments Left is greater
      than zero, the destination address is the last address
      (Address[n]) in the Routing header.



McCann, Deering & Mogul     Standards Track                     [Page 7]

RFC 1981              Path MTU Discovery for IPv6            August 1996


   The node then uses the value in the MTU field in the Packet Too Big
   message as a tentative PMTU value, and compares the tentative PMTU to
   the existing PMTU.  If the tentative PMTU is less than the existing
   PMTU estimate, the tentative PMTU replaces the existing PMTU as the
   PMTU value for the path.

   The packetization layers must be notified about decreases in the
   PMTU.  Any packetization layer instance (for example, a TCP
   connection) that is actively using the path must be notified if the
   PMTU estimate is decreased.

      Note: even if the Packet Too Big message contains an Original
      Packet Header that refers to a UDP packet, the TCP layer must be
      notified if any of its connections use the given path.

   Also, the instance that sent the packet that elicited the Packet Too
   Big message should be notified that its packet has been dropped, even
   if the PMTU estimate has not changed, so that it may retransmit the
   dropped data.

      Note: An implementation can avoid the use of an asynchronous
      notification mechanism for PMTU decreases by postponing
      notification until the next attempt to send a packet larger than
      the PMTU estimate.  In this approach, when an attempt is made to
      SEND a packet that is larger than the PMTU estimate, the SEND
      function should fail and return a suitable error indication.  This
      approach may be more suitable to a connectionless packetization
      layer (such as one using UDP), which (in some implementations) may
      be hard to "notify" from the ICMP layer.  In this case, the normal
      timeout-based retransmission mechanisms would be used to recover
      from the dropped packets.

   It is important to understand that the notification of the
   packetization layer instances using the path about the change in the
   PMTU is distinct from the notification of a specific instance that a
   packet has been dropped.  The latter should be done as soon as
   practical (i.e., asynchronously from the point of view of the
   packetization layer instance), while the former may be delayed until
   a packetization layer instance wants to create a packet.
   Retransmission should be done for only for those packets that are
   known to be dropped, as indicated by a Packet Too Big message.

5.3. Purging stale PMTU information



   Internetwork topology is dynamic; routes change over time.  While the
   local representation of a path may remain constant, the actual
   path(s) in use may change.  Thus, PMTU information cached by a node
   can become stale.



McCann, Deering & Mogul     Standards Track                     [Page 8]

RFC 1981              Path MTU Discovery for IPv6            August 1996


   If the stale PMTU value is too large, this will be discovered almost
   immediately once a large enough packet is sent on the path.  No such
   mechanism exists for realizing that a stale PMTU value is too small,
   so an implementation should "age" cached values.  When a PMTU value
   has not been decreased for a while (on the order of 10 minutes), the
   PMTU estimate should be set to the MTU of the first-hop link, and the
   packetization layers should be notified of the change.  This will
   cause the complete Path MTU Discovery process to take place again.

      Note: an implementation should provide a means for changing the
      timeout duration, including setting it to "infinity".  For
      example, nodes attached to an FDDI link which is then attached to
      the rest of the Internet via a small MTU serial line are never
      going to discover a new non-local PMTU, so they should not have to
      put up with dropped packets every 10 minutes.

   An upper layer must not retransmit data in response to an increase in
   the PMTU estimate, since this increase never comes in response to an
   indication of a dropped packet.

   One approach to implementing PMTU aging is to associate a timestamp
   field with a PMTU value.  This field is initialized to a "reserved"
   value, indicating that the PMTU is equal to the MTU of the first hop
   link.  Whenever the PMTU is decreased in response to a Packet Too Big
   message, the timestamp is set to the current time.

   Once a minute, a timer-driven procedure runs through all cached PMTU
   values, and for each PMTU whose timestamp is not "reserved" and is
   older than the timeout interval:

   - The PMTU estimate is set to the MTU of the first hop link.

   - The timestamp is set to the "reserved" value.

   - Packetization layers using this path are notified of the increase.

5.4. TCP layer actions



   The TCP layer must track the PMTU for the path(s) in use by a
   connection; it should not send segments that would result in packets
   larger than the PMTU.  A simple implementation could ask the IP layer
   for this value each time it created a new segment, but this could be
   inefficient.  Moreover, TCP implementations that follow the "slow-
   start" congestion-avoidance algorithm [CONG] typically calculate and
   cache several other values derived from the PMTU.  It may be simpler
   to receive asynchronous notification when the PMTU changes, so that
   these variables may be updated.




McCann, Deering & Mogul     Standards Track                     [Page 9]

RFC 1981              Path MTU Discovery for IPv6            August 1996


   A TCP implementation must also store the MSS value received from its
   peer, and must not send any segment larger than this MSS, regardless
   of the PMTU.  In 4.xBSD-derived implementations, this may require
   adding an additional field to the TCP state record.

   The value sent in the TCP MSS option is independent of the PMTU.
   This MSS option value is used by the other end of the connection,
   which may be using an unrelated PMTU value.  See [IPv6-SPEC] sections
   "Packet Size Issues" and "Maximum Upper-Layer Payload Size" for
   information on selecting a value for the TCP MSS option.

   When a Packet Too Big message is received, it implies that a packet
   was dropped by the node that sent the ICMP message.  It is sufficient
   to treat this as any other dropped segment, and wait until the
   retransmission timer expires to cause retransmission of the segment.
   If the Path MTU Discovery process requires several steps to find the
   PMTU of the full path, this could delay the connection by many
   round-trip times.

   Alternatively, the retransmission could be done in immediate response
   to a notification that the Path MTU has changed, but only for the
   specific connection specified by the Packet Too Big message.  The
   packet size used in the retransmission should be no larger than the
   new PMTU.

      Note: A packetization layer must not retransmit in response to
      every Packet Too Big message, since a burst of several oversized
      segments will give rise to several such messages and hence several
      retransmissions of the same data.  If the new estimated PMTU is
      still wrong, the process repeats, and there is an exponential
      growth in the number of superfluous segments sent.

      This means that the TCP layer must be able to recognize when a
      Packet Too Big notification actually decreases the PMTU that it
      has already used to send a packet on the given connection, and
      should ignore any other notifications.

   Many TCP implementations incorporate "congestion avoidance" and
   "slow-start" algorithms to improve performance [CONG].  Unlike a
   retransmission caused by a TCP retransmission timeout, a
   retransmission caused by a Packet Too Big message should not change
   the congestion window.  It should, however, trigger the slow-start
   mechanism (i.e., only one segment should be retransmitted until
   acknowledgements begin to arrive again).

   TCP performance can be reduced if the sender's maximum window size is
   not an exact multiple of the segment size in use (this is not the
   congestion window size, which is always a multiple of the segment



McCann, Deering & Mogul     Standards Track                    [Page 10]

RFC 1981              Path MTU Discovery for IPv6            August 1996


   size).  In many systems (such as those derived from 4.2BSD), the
   segment size is often set to 1024 octets, and the maximum window size
   (the "send space") is usually a multiple of 1024 octets, so the
   proper relationship holds by default.  If Path MTU Discovery is used,
   however, the segment size may not be a submultiple of the send space,
   and it may change during a connection; this means that the TCP layer
   may need to change the transmission window size when Path MTU
   Discovery changes the PMTU value.  The maximum window size should be
   set to the greatest multiple of the segment size that is less than or
   equal to the sender's buffer space size.

5.5. Issues for other transport protocols



   Some transport protocols (such as ISO TP4 [ISOTP]) are not allowed to
   repacketize when doing a retransmission.  That is, once an attempt is
   made to transmit a segment of a certain size, the transport cannot
   split the contents of the segment into smaller segments for
   retransmission.  In such a case, the original segment can be
   fragmented by the IP layer during retransmission.  Subsequent
   segments, when transmitted for the first time, should be no larger
   than allowed by the Path MTU.

   The Sun Network File System (NFS) uses a Remote Procedure Call (RPC)
   protocol [RPC] that, when used over UDP, in many cases will generate
   payloads that must be fragmented even for the first-hop link.  This
   might improve performance in certain cases, but it is known to cause
   reliability and performance problems, especially when the client and
   server are separated by routers.

   It is recommended that NFS implementations use Path MTU Discovery
   whenever routers are involved.  Most NFS implementations allow the
   RPC datagram size to be changed at mount-time (indirectly, by
   changing the effective file system block size), but might require
   some modification to support changes later on.

   Also, since a single NFS operation cannot be split across several UDP
   datagrams, certain operations (primarily, those operating on file
   names and directories) require a minimum payload size that if sent in
   a single packet would exceed the PMTU.  NFS implementations should
   not reduce the payload size below this threshold, even if Path MTU
   Discovery suggests a lower value.  In this case the payload will be
   fragmented by the IP layer.









McCann, Deering & Mogul     Standards Track                    [Page 11]

RFC 1981              Path MTU Discovery for IPv6            August 1996


5.6. Management interface



   It is suggested that an implementation provide a way for a system
   utility program to:

   - Specify that Path MTU Discovery not be done on a given path.

   - Change the PMTU value associated with a given path.

   The former can be accomplished by associating a flag with the path;
   when a packet is sent on a path with this flag set, the IP layer does
   not send packets larger than the IPv6 minimum link MTU.

   These features might be used to work around an anomalous situation,
   or by a routing protocol implementation that is able to obtain Path
   MTU values.

   The implementation should also provide a way to change the timeout
   period for aging stale PMTU information.

6. Security Considerations



   This Path MTU Discovery mechanism makes possible two denial-of-
   service attacks, both based on a malicious party sending false Packet
   Too Big messages to a node.

   In the first attack, the false message indicates a PMTU much smaller
   than reality.  This should not entirely stop data flow, since the
   victim node should never set its PMTU estimate below the IPv6 minimum
   link MTU.  It will, however, result in suboptimal performance.

   In the second attack, the false message indicates a PMTU larger than
   reality.  If believed, this could cause temporary blockage as the
   victim sends packets that will be dropped by some router.  Within one
   round-trip time, the node would discover its mistake (receiving
   Packet Too Big messages from that router), but frequent repetition of
   this attack could cause lots of packets to be dropped.  A node,
   however, should never raise its estimate of the PMTU based on a
   Packet Too Big message, so should not be vulnerable to this attack.

   A malicious party could also cause problems if it could stop a victim
   from receiving legitimate Packet Too Big messages, but in this case
   there are simpler denial-of-service attacks available.








McCann, Deering & Mogul     Standards Track                    [Page 12]

RFC 1981              Path MTU Discovery for IPv6            August 1996


Acknowledgements



   We would like to acknowledge the authors of and contributors to
   [RFC-1191], from which the majority of this document was derived.  We
   would also like to acknowledge the members of the IPng working group
   for their careful review and constructive criticisms.













































McCann, Deering & Mogul     Standards Track                    [Page 13]

RFC 1981              Path MTU Discovery for IPv6            August 1996


Appendix A - Comparison to RFC 1191

   This document is based in large part on RFC 1191, which describes
   Path MTU Discovery for IPv4.  Certain portions of RFC 1191 were not
   needed in this document:

   router specification    - Packet Too Big messages and corresponding
                             router behavior are defined in [ICMPv6]

   Don't Fragment bit      - there is no DF bit in IPv6 packets

   TCP MSS discussion      - selecting a value to send in the TCP MSS
                             option is discussed in [IPv6-SPEC]

   old-style messages      - all Packet Too Big messages report the
                             MTU of the constricting link

   MTU plateau tables      - not needed because there are no old-style
                             messages

References

   [CONG]      Van Jacobson.  Congestion Avoidance and Control.  Proc.
               SIGCOMM '88 Symposium on Communications Architectures and
               Protocols, pages 314-329.  Stanford, CA, August, 1988.

   [FRAG]      C. Kent and J. Mogul.  Fragmentation Considered Harmful.
               In Proc. SIGCOMM '87 Workshop on Frontiers in Computer
               Communications Technology.  August, 1987.

   [ICMPv6]    Conta, A., and S. Deering, "Internet Control Message
               Protocol (ICMPv6) for the Internet Protocol Version 6
               (IPv6) Specification", RFC 1885, December 1995.

   [IPv6-SPEC] Deering, S., and R. Hinden, "Internet Protocol, Version
               6 (IPv6) Specification", RFC 1883, December 1995.

   [ISOTP]     ISO.  ISO Transport Protocol Specification: ISO DP 8073.
               RFC 905, SRI Network Information Center, April, 1984.

   [ND]        Narten, T., Nordmark, E., and W. Simpson, "Neighbor
               Discovery for IP Version 6 (IPv6)", Work in Progress.

   [RFC-1191]  Mogul, J., and S. Deering, "Path MTU Discovery",
               RFC 1191, November 1990.






McCann, Deering & Mogul     Standards Track                    [Page 14]

RFC 1981              Path MTU Discovery for IPv6            August 1996


   [RPC]       Sun Microsystems, Inc., "RPC: Remote Procedure Call
               Protocol", RFC 1057, SRI Network Information Center,
               June, 1988.

Authors' Addresses



   Jack McCann
   Digital Equipment Corporation
   110 Spitbrook Road, ZKO3-3/U14
   Nashua, NH 03062
   Phone: +1 603 881 2608

   Fax:   +1 603 881 0120
   Email: mccann@zk3.dec.com


   Stephen E. Deering
   Xerox Palo Alto Research Center
   3333 Coyote Hill Road
   Palo Alto, CA 94304
   Phone: +1 415 812 4839

   Fax:   +1 415 812 4471
   EMail: deering@parc.xerox.com


   Jeffrey Mogul
   Digital Equipment Corporation Western Research Laboratory
   250 University Avenue
   Palo Alto, CA 94301
   Phone: +1 415 617 3304

   EMail: mogul@pa.dec.com


















McCann, Deering & Mogul     Standards Track                    [Page 15]