Network Working Group                                        A. McKenzie
Request for Comments: 1110                                       BBN STC
                                                            August 1989


               A Problem with the TCP Big Window Option

Status of this Memo

  This memo comments on the TCP Big Window option described in RFC
  1106.  Distribution of this memo is unlimited.

Abstract

  The TCP Big Window option discussed in RFC 1106 will not work
  properly in an Internet environment which has both a high bandwidth *
  delay product and the possibility of disordering and duplicating
  packets.  In such networks, the window size must not be increased
  without a similar increase in the sequence number space.  Therefore,
  a different approach to big windows should be taken in the Internet.

Discussion

  TCP was designed to work in a packet store-and-forward environment
  characterized by the possibility of packet loss, packet disordering,
  and packet duplication.  Packet loss can occur, for example, by a
  congested network element discarding a packet.  Packet disordering
  can occur, for example, by packets of a TCP connection being
  arbitrarily transmitted partially over a low bandwidth terrestrial
  path and partially over a high bandwidth satellite path.  Packet
  duplication can occur, for example, when two directly-connected
  network elements use a reliable link protocol and the link goes down
  after the receiver correctly receives a packet but before the
  transmitter receives an acknowledgement for the packet; the
  transmitter and receiver now each take responsibility for attempting
  to deliver the same packet to its ultimate destination.

  TCP has the task of recreating at the destination an exact copy of
  the data stream generated at the source, in the same order and with
  no gaps or duplicates.  The mechanism used to accomplish this task is
  to assign a "unique" sequence number to each byte of data at its
  source, and to sort the bytes at the destination according to the
  sequence number.  The sorting operation corrects any disordering.  An
  acknowledgement, timeout, and retransmission scheme corrects for data
  loss.  The uniqueness of the sequence number corrects for data
  duplication.

  As a practical matter, however, the sequence number is not unique; it



McKenzie                                                        [Page 1]

RFC 1110           Comments on TCP Big Window Option         August 1989


  is contained in a 32-bit field and therefore "wraps around" after the
  transmission of 2**32 bytes of data.  Two additional mechanisms are
  used to insure the effective uniqueness of sequence numbers; these
  are the TCP transmission window and bounds on packet lifetime within
  the Internet, including the IP Time-to-Live (TTL).  The transmission
  window specifies the maximum number of bytes which may be sent by the
  source in one source-destination roundtrip time.  Since the TCP
  transmission window is specified by 16 bits, which is 1/65536 of the
  sequence number space, a sequence number will not be reused (used to
  number another byte) for 65,536 roundtrip times.  So long as the
  combination of gateway action on the IP TTL and holding times within
  the individual networks which interconnect the gateways do not allow
  a packet's lifetime to exceed 65,536 roundtrip times, each sequence
  number is effectively unique.  It was believed by the TCP designers
  that the networks and gateways forming the internet would meet this
  constraint, and such has been the case.

  The proposed TCP Big Window option, as described in RFC 1106, expands
  the size of the window specification to 30 bits, while leaving the
  sequence number space unchanged.  Thus, a sequence number can be
  reused after 4 roundtrip times.  Further, the Nak option allows a
  packet to be retransmitted (i.e., potentially duplicated) by the
  source after only one roundtrip time.  Thus, if a packet becomes
  "lost" in the Internet for only about 5 roundtrip times it may be
  delivered when its sequence number again lies within the window,
  albeit a later cycle of the window.  In this case, TCP will not
  necessarily recreate at the destination an exact copy of the data
  stream generated at the source; it may replace some data with earlier
  data.

  Of course, the problem described above results from the storage of
  the "lost" packet within the net, and its subsequent out-of-order
  delivery.  RFC 1106 seems to describe use of the proposed options in
  an isolated satellite network.  We may hypothesize that this network
  is memoryless, and thus cannot deliver packets out of order; it
  either delivers a packet in order or loses it.  If this is the case,
  then there is no problem with the proposed options.  The Internet,
  however, can deliver packets out of order, and this will likely
  continue to be true even if gigabit links become part of the
  Internet.  Therefore, the approach described in RFC 1106 cannot be
  adopted for general Internet use.










McKenzie                                                        [Page 2]

RFC 1110           Comments on TCP Big Window Option         August 1989


Author's Address

  Alex McKenzie
  Bolt Beranek and Newman Inc.
  10 Moulton Street
  Cambridge, MA 02238

  Phone: (617) 873-2962

  EMail: [email protected]









































McKenzie                                                        [Page 3]


You are viewing proxied material from gopher.fnord.one. 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.