Network Working Group                                          G. Malkin
Request For Comments: 1868                                Xylogics, Inc.
Category: Experimental                                     November 1995


                        ARP Extension - UNARP

Status of this Memo

  This memo defines an Experimental Protocol for the Internet
  community.  This memo does not specify an Internet standard of any
  kind.  Discussion and suggestions for improvement are requested.
  Distribution of this memo is unlimited.

Abstract

  The Address Resolution Protocol allows an IP node to determine the
  hardware (datalink) address of a neighboring node on a broadcast
  network.  The protocol depends on timers to age away old ARP entries.
  This document specifies a trivial modification to the ARP mechanism,
  not the packet format, which allows a node to announce that it is
  leaving the network and that all other nodes should modify their ARP
  tables accordingly.

Acknowledgements

  Thanks to James Carlson/Xylogics for reviewing this document and
  proposing the backwards compatibility mechanism.

1. Introduction

  The primary purpose of the Address Resolution Protocol, as defined in
  [1], is to determine a node's hardware address based on its network
  address (protocol address in ARPspeak).  The ARP protocol
  specifically states that nodes should not periodically advertise
  their existence for two reasons: first, this would generate a lot of
  network traffic and table maintenance overhead; second, it is highly
  unlikely that all nodes will need to communicate to all other nodes.
  Since a node does not advertise its existence, neither does it
  advertise its imminent departure.  This is not a serious problem
  since most ARP implementations maintain timers to age away old
  entries, and departing nodes seldom depart gracefully in any case.

  Over time, an additional use has been found for ARP: Proxy ARP.
  While there are those who believe Proxy ARP is an evil thing, it does
  serve a purpose; that is, it allows for communication in ways never
  considered in the original IP architecture.  For example, allows
  dial-in hosts to connect to a network without consuming a large



Malkin                        Experimental                      [Page 1]

RFC 1868                         UNARP                     November 1995


  amount of the IP address space (i.e., all of the hosts contain
  addresses on the same subnet, even though they are not directly
  attached to the physical network associated with that subnet address.
  It is this use of Proxy ARP which produces the problem addressed by
  this document.

2. The Problem

  Consider the following topology:

                   +--------+
                   | Host A |
                   +--------+
                       |
     ======================================== LAN
         |                             |
     +--------+                    +--------+
     |  CS1   |   comm. servers    |  CS2   |
     +--------+                    +--------+
       |    |                        |    |
      +-+  +-+                      +-+  +-+
      | |  | |       modems         | |  | |
      +-+  +-+                      +-+  +-+

  Assume that all of the modems are on the same rotary; that is, when a
  remote host dials in, it may be assigned a modem on either of the
  communication servers.  Further assume that all of the remote hosts'
  IP addresses have the same subnet address as the servers and Host A,
  this in order to conserve address space.

  To begin, a remote host dials into CS1 and attempts to communicate
  with Host A.  Host A will assume, based on the subnet mask, that the
  remote host is actually attached to the LAN and will issue an ARP
  Request to determine its hardware address.  Naturally, the remote
  host will not hear this request.  CS1, knowing this, will respond in
  the remote host's place with its own hardware address.  Host A, on
  receiving the ARP Reply, will then communicate with the remote host,
  transparently through CS1.  So far everything is just fine.

  Now, the remote host disconnects and, before Host A can age its ARP
  cache, reconnects through CS2.  Herein lies the problem.  Whenever
  Host A attempts to send a packet to the remote host, it will send it
  to CS1 because it cannot know that its ARP cache entry is invalid.
  If, when the remote host disconnects, the server to which it was
  attached could inform other nodes on the LAN that the protocol
  address/hardware address mapping was no longer valid, the problem
  would not occur.




Malkin                        Experimental                      [Page 2]

RFC 1868                         UNARP                     November 1995


3. The Solution

  When a server, as described above, disconnects from a remote host for
  which it has responded to a Proxy ARP, it broadcasts an UNARP.  An
  UNARP is an unsolicited ARP Reply with the following field values:

     Hardware Address Space       as appropriate
     Protocol Address Space       0x800 (IP)
     Hardware Address Length      0 (see Backwards Compatibility)
     Protocol Address Length      4 (length of an IP address)
     Opcode                       2 (Reply)
     Source Hardware Address      Not Included
     Source Protocol Address      IP address of detaching host
     Target Hardware Address      Not Included
     Target Protocol Address      255.255.255.255 (IP broadcast)

     NOTE: this is a 16-byte packet (not including MAC header)

  On receiving an UNARP, a node deletes the ARP cache entry associated
  with the IP address.

  It is not strictly necessary that a server keep state information
  about whether or not it has actually sent a Proxy ARP Reply; it would
  be sufficient if a server always sends an UNARP when a remote host
  disconnects.

  Of course, there is no reason why a host which gracefully detaches
  from a LAN cannot also send an UNARP for itself.  This would be
  especially useful if, upon re-attaching, it might have a different
  hardware address.

4. Backwards Compatibility

  The modifications to support UNARP are trivial, so there is every
  expectation that it will be widely supported.  Of course, there will
  be a period of time during which nodes which support UNARP will
  coexist with nodes which do not support UNARP.  To prevent
  unenlightened nodes from adding spurious ARP cache entries with
  hardware addresses of zero, UNARP packets specify a hardware address
  length of zero.  This should be rejected by nodes which do not
  support UNARP.  As a consequence of this, the source and target
  hardware address fields do not exist in UNARP packets (as previously
  described).

  It is recommended that implementors include a configuration switch to
  disable UNARP in the event that some vendor's ARP implementation
  might take offense at the abbreviated UNARP packet format.




Malkin                        Experimental                      [Page 3]

RFC 1868                         UNARP                     November 1995


5. Security Considerations

  Security issues are not discussed in this memo.

References

  [1] Plummer, D., "An Ethernet Address Resolution Protocol", STD 37,
      RFC 826, MIT, November 1982.

Author's Address

  Gary Scott Malkin
  Xylogics, Inc.
  53 Third Avenue
  Burlington, MA  01803

  Phone:  (617) 272-8140
  EMail:  [email protected]

































Malkin                        Experimental                      [Page 4]


You are viewing proxied material from gopher.rbfh.de. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.