Network Working Group                                     David D. Clark
Request for Comments: 969                                Mark L. Lambert
                                                             Lixia Zhang
                                M. I. T. Laboratory for Computer Science
                                                           December 1985

                 NETBLT: A Bulk Data Transfer Protocol


1. STATUS OF THIS MEMO

   This RFC suggests a proposed protocol for the ARPA-Internet
   community, and requests discussion and suggestions for improvements.
   This is a preliminary discussion of the NETBLT protocol.  It is
   published for discussion and comment, and does not constitute a
   standard.  As the proposal may change, implementation of this
   document is not advised.  Distribution of this memo is unlimited.

2. INTRODUCTION

   NETBLT (Network Block Transfer) is a transport level protocol
   intended for the rapid transfer of a large quantity of data between
   computers. It provides a transfer that is reliable and flow
   controlled, and is structured to provide maximum throughput over a
   wide variety of networks.

   The protocol works by opening a connection between two clients the
   sender and the receiver), transferring the data in a series of large
   data aggregates called buffers, and then closing the connection.
   Because the amount of data to be transferred can be arbitrarily
   large, the client is not required to provide at once all the data to
   the protocol module.  Instead, the data is provided by the client in
   buffers.  The NETBLT layer transfers each buffer as a sequence of
   packets, but since each buffer is composed of a large number of
   packets, the per-buffer interaction between NETBLT and its client is
   far more efficient than a per-packet interaction would be.

   In its simplest form, a NETBLT transfer works as follows.  The
   sending client loads a buffer of data and calls down to the NETBLT
   layer to transfer it.  The NETBLT layer breaks the buffer up into
   packets and sends these packets across the network in Internet
   datagrams.  The receiving NETBLT layer loads these packets into a
   matching buffer provided by the receiving client.  When the last
   packet in the buffer has been transmitted, the receiving NETBLT
   checks to see that all packets in that buffer have arrived.  If some
   packets are missing, the receiving NETBLT requests that they be
   resent.  When the buffer has been completely transmitted, the
   receiving client is notified by its NETBLT layer.  The receiving
   client disposes of the buffer and provides a new buffer to receive
   more data.  The receiving NETBLT notifies the sender that the buffer
   arrived, and the sender prepares and sends the next buffer in the

   same manner.  This continues until all buffers have been sent, at
   which time the sender notifies the receiver that the transmission has
   been completed.  The connection is then closed.

   As described above, the NETBLT protocol is "lock-step"; action is
   halted after a buffer is transmitted, and begins again after
   confirmation is received from the receiver of data.  NETBLT provides
   for multiple buffering, in which several buffers can be transmitted
   concurrently.  Multiple buffering makes packet flow essentially
   continuous and can improve performance markedly.

   The remainder of this document describes NETBLT in detail.  The next
   sections describe the philosophy behind a number of protocol
   features: packetization, flow control, reliability, and connection
   management. The final sections describe the protocol format.

3. BUFFERS AND PACKETS

   NETBLT is designed to permit transfer of an essentially arbitrary
   amount of data between two clients.  During connection setup the
   sending NETBLT can optionally inform the receiving NETBLT of the
   transfer size; the maximum transfer length is imposed by the field
   width, and is 2**32 bytes.  This limit should permit any practical
   application.  The transfer size parameter is for the use of the
   receiving client; the receiving NETBLT makes no use of it.  A NETBLT
   receiver accepts data until told by the sender that the transfer is
   complete.

   The data to be sent must be broken up into buffers by the client.
   Each buffer must be the same size, save for the last buffer.  During
   connection setup, the sending and receiving NETBLTs negotiate the
   buffer size, based on limits provided by the clients.  Buffer sizes
   are in bytes only; the client is responsible for breaking up data
   into buffers on byte boundaries.

   NETBLT has been designed and should be implemented to work with
   buffers of arbitrary size.  The only fundamental limitation on buffer
   size should be the amount of memory available to the client.  Buffers
   should be as large as possible since this minimizes the number of
   buffer transmissions and therefore improves performance.

   NETBLT is designed to require a minimum of its own memory, allowing
   the client to allocate as much memory as possible for buffer storage.
   In particular, NETBLT does not keep buffer copies for retransmission
   purposes.  Instead, data to be retransmitted is recopied directly

   from the client buffer.  This does mean that the client cannot
   release buffer storage piece by piece as the buffer is sent, but this
   has not proved a problem in preliminary NETBLT implementations.

   Buffers are broken down by the NETBLT layer into sequences of DATA
   packets.  As with the buffer size, the packet size is negotiated
   between the sending and receiving NETBLTs during connection setup.
   Unlike buffer size, packet size is visible only to the NETBLT layer.

   All DATA packets save the last packet in a buffer must be the same
   size.  Packets should be as large as possible, since in most cases
   (including the preliminary protocol implementation) performance is
   directly related to packet size.  At the same time, the packets
   should not be so large as to cause Internet fragmentation, since this
   normally causes performance degrada- tion.

   All buffers save the last buffer must be the same size; obviously the
   last buffer can be any size required to complete the transfer. Since
   the receiving NETBLT does not know the transfer size in advance, it
   needs some way of identifying the last packet in each buffer.  For
   this reason, the last packet of every buffer is not a DATA packet but
   rather an LDATA packet.  DATA and LDATA packets are identical save
   for the packet type.

4. FLOW CONTROL

   NETBLT uses two strategies for flow control, one internal and one at
   the client level.

   The sending and receiving NETBLTs transmit data in buffers; client
   flow control is therefore at a buffer level.  Before a buffer can be
   transmitted, NETBLT confirms that both clients have set up matching
   buffers, that one is ready to send data, and that the other is ready
   to receive data.  Either client can therefore control the flow of
   data by not providing a new buffer.  Clients cannot stop a buffer
   transfer while it is in progress.

   Since buffers can be quite large, there has to be another method for
   flow control that is used during a buffer transfer.  The NETBLT layer
   provides this form of flow control.

   There are several flow control problems that could arise while a
   buffer is being transmitted.  If the sending NETBLT is transferring
   data faster than the receiving NETBLT can process it, the receiver's
   ability to buffer unprocessed packets could be overflowed, causing
   packets to be lost.  Similarly, a slow gateway or intermediate
   network could cause packets to collect and overflow network packet

   buffer space.  Packets will then be lost within the network,
   degrading performance.  This problem is particularly acute for NETBLT
   because NETBLT buffers will generally be quite large, and therefore
   composed of many packets.

   A traditional solution to packet flow control is a window system, in
   which the sending end is permitted to send only a certain number of
   packets at a time.  Unfortunately, flow control using windows tends
   to result in low throughput.  Windows must be kept small in order to
   avoid overflowing hosts and gateways, and cannot easily be updated,
   since an end-to-end exchange is required for each change.

   To permit high throughput over a variety of networks and gateways of
   differing speeds, NETBLT uses a novel flow control ethod: rate
   control.  The transmission rate is negotiated by the sending and
   receiving NETBLTs during connection setup and after each buffer
   transmission.  The sender uses timers, rather than messages from the
   receiver, to maintain the negotiated rate.

   In its simplest form, rate control specifies a minimum time period
   per packet transmission.  This can cause performance problems for
   several reasons: the transmission time for a single packet is very
   small, frequently smaller than the granularity of the timing
   mechanism.  Also, the overhead required to maintain timing mechanisms
   on a per packet basis is relatively high, which degrades performance.

   The solution is to control the transmission rate of groups of
   packets, rather than single packets.  The sender transmits a burst of
   packets over negotiated interval, then sends another burst.  In this
   way, the overhead decreases by a factor of the burst size, and the
   per-burst transmission rate is large enough that timing mechanisms
   will work properly.  The NETBLT's rate control therefore has two
   parts, a burst size and a burst rate, with (burst size)/(burst rate)
   equal to the average transmission rate per packet.

   The burst size and burst rate should be based not only on the packet
   transmission and processing speed which each end can handle, but also
   on the capacities of those gateways and networks intermediate to the
   transfer.  Following are some intuitive values for packet size,
   buffer size, burst size, and burst rate.

   Packet sizes can be as small as 128 bytes.  Performance with packets
   this small is almost always bad, because of the high per-packet
   processing overhead.  Even the default Internet Protocol packet size
   of 576 bytes is barely big enough for adequate performance.  Most

   networks do not support packet sizes much larger than one or two
   thousand bytes, and packets of this size can also get fragmented when
   traveling over intermediate networks, degrading performance.

   The size of a NETBLT buffer is limited only by the amount of memory
   available to a client.  Theoretically, buffers of 100K bytes or more
   are possible.  This would mean the transmission of 50 to 100 packets
   per buffer.

   The burst size and burst rate are obviously very machine dependent.
   There is a certain amount of transmission overhead in the sending and
   receiving machines associated with maintaining timers and scheduling
   processes.  This overhead can be minimized by sending packets in
   large bursts.  There are also limitations imposed on the burst size
   by the number of available packet buffers.  On most modern operating
   systems, a burst size of between five and ten packets should reduce
   the overhead to an acceptable level.  In fact, a preliminary NETBLT
   implementation for the IBM PC/AT sends packets in bursts of five.  It
   could send more, but is limited by available memory.

   The burst rate is in part determined by the granularity of the
   sender's timing mechanism, and in part by the processing speed of the
   receiver and any intermediate gateways.  It is also directly related
   to the burst size.  Burst rates from 60 to 100 milliseconds have been
   tried on the preliminary NETBLT implementation with good results
   within a single local-area network.  This value clearly depends on
   the network bandwidth and packet buffering available.

   All NETBLT flow control parameters (packet size, buffer size, burst
   size, and burst rate) are negotiated during connection setup.  The
   negotiation process is the same for all parameters.  The client
   initiating the connection (the active end) proposes and sends a set
   of values for each parameter with its open connection request.  The
   other client (the passive end) compares these values with the
   highest-performance values it can support.  The passive end can then
   modify any of the parameters only by making them more restrictive.
   The modified parameters are then sent back to the active end in the
   response message.  In addition, the burst size and burst rate can be
   re-negotiated after each buffer transmission to adjust the transfer
   rate according to the performance observed from transferring the
   previous buffer.  The receiving end sends a pair of burst size and
   burst rate values in the OK message.  The sender compares these
   values with the values it can support.  Again, it may then modify any
   of the parameters only by making them more restrictive.  The modified
   parameters are then communicated to the receiver in a NULL-ACK
   packet, described later.

   Obviously each of the parameters depend on many factors-- gateway and
   host processing speeds, available memory, timer granularity--some of
   which cannot be checked by either client.  Each client must therefore
   try to make as best a guess as it can, tuning for performance on
   subsequent transfers.

5. RELIABILITY

   Each NETBLT transfer has three stages, connection setup, data
   transfer, and connection close.  Each stage must be completed
   reliably; methods for doing this are described below.

   5.1. Connection Setup

      A NETBLT connection is set up by an exchange of two packets
      between the active client and the passive client.  Note that
      either client can send or receive data; the words "active" and
      "passive" are only used to differentiate the client initiating the
      connection process from the client responding to the connection
      request.  The first packet sent is an OPEN packet; the passive end
      acknowledges the OPEN packet by sending a RESPONSE packet.  After
      these two packets have been exchanged, the transfer can begin.

      As discussed in the previous section, the OPEN and RESPONSE
      packets are used to negotiate flow control parameters.  Other
      parameters used in the transfer of data are also negotiated.
      These parameters are (1) the maximum number of buffers that can be
      sending at any one time (this permits multiple buffering and
      higher throughput) and (2) whether or not DATA/LDATA packet data
      will be checksummed.  NETBLT automatically checksums all
      non-DATA/LDATA packets.  If the negotiated checksum flag is set to
      TRUE (1), both the header and the data of a DATA/LDATA packet are
      checksummed; if set to FALSE (0), only the header is checksummed.
      NETBLT uses the same checksumming algorithm as TCP uses.

      Finally, each end transmits its death-timeout value in either the
      OPEN or the RESPONSE packet.  The death-timeout value will be used
      to determine the frequency with which to send KEEPALIVE packets
      during idle periods of an opened connection (death timers and
      KEEPALIVE packets are described in the following section).

      The active end specifies a passive client through a
      client-specific "well-known" 16 bit port number on which the
      passive end listens.  The active end identifies itself through a
      32 bit Internet address and a 16 bit port number.

      In order to allow the active and passive ends to communicate

      miscellaneous useful information, an unstructured, variable-
      length field is provided in OPEN and RESPONSE messages for an
      client-specific information that may be required.

      Recovery for lost OPEN and RESPONSE packets is provided by the use
      of timers.  The active end sets a timer when it sends an OPEN
      packet. When the timer expires, another OPEN packet is sent, until
      some pre-determined maximum number of OPEN packets have been sent.
      A similar scheme is used for the passive end when it sends a
      RESPONSE packet.  When a RESPONSE packet is received by the active
      end, it clears its timer.  The passive end's timer is cleared
      either by receipt of a GO or a DATA packet, as described in the
      section on data transfer.

      To prevent duplication of OPEN and RESPONSE packets, the OPEN
      packet contains a 32 bit connection unique ID that must be
      returned in the RESPONSE packet.  This prevents the initiator from
      confusing the response to the current request with the response to
      an earlier connection request (there can only be one connection
      between any two ports).  Any OPEN or RESPONSE packet with a
      destination port matching that of an open connection has its
      unique ID checked.  A matching unique ID implies a duplicate
      packet, and the packet is ignored.  A non-matching unique ID must
      be treated as an attempt to open a second connection between the
      same port pair and must be rejected by sending an ABORT message.

   5.2. Data Transfer

      The simplest model of data transfer proceeds as follows.  The
      sending client sets up a buffer full of data.  The receiving
      NETBLT sends a GO message inside a CONTROL packet to the sender,
      signifying that it too has set up a buffer and is ready to receive
      data into it. Once the GO message has been received, the sender
      transmits the buffer as a series of DATA packets followed by an
      LDATA packet.  When the last packet in the buffer has been
      received, the receiver sends a RESEND message inside a CONTROL
      packet containing a list of packets that were not received.  The
      sender resends these packets.  This process continues until there
      are no missing packets, at which time the receiver sends an OK
      message inside a CONTROL packet to the sender, sets up another
      buffer to receive data and sends another GO message.  The sender,
      having received the OK message, sets up another buffer, waits for
      the GO message, and repeats the process.

      There are several obvious flaws with this scheme.  First, if the
      LDATA packet is lost, how does the receiver know when the buffer
      has been transmitted?  Second, what if the GO, OK, or RESEND

      messages are lost?  The sender cannot act on a packet it has not
      received, so the protocol will hang.  Solutions for each of these
      problems are presented below, and are based on two kinds of
      timers, a data timer and a control timer.

      NETBLT solves the LDATA packet loss problem by using a data timer
      at the receiving end.  When the first DATA packet in a buffer
      arrives, the receiving NETBLT sets its data timer; at the same
      time, it clears its control timer, described below.  If the data
      timer expires, the receiving end assumes the buffer has been
      transmitted and all missing packets lost.  It then sends a RESEND
      message containing a list of the missing packets.

      NETBLT solves the second problem, that of missing OK, GO, and
      RESEND messages, through use of a control timer.  The receiver can
      send one or more control messages (OK, GO, or RESEND) within a
      single CONTROL packet.  Whenever the receiver sends a control
      packet, it sets a control timer (at the same time it clears its
      data timer, if one has been set).

      The control timer is cleared as follows: Each control message
      includes a sequence number which starts at one and increases by
      one for each control message sent.  The sending NETBLT checks the
      sequence number of every incoming control message against all
      other sequence numbers it has received.  It stores the highest
      sequence number below which all other received sequence numbers
      are consecutive, and returns this number in every packet flowing
      back to the receiver.  The receiver is permitted to clear the
      control timer of every packet with a sequence number equal to or
      lower than the sequence number returned by the sender.

      Ideally, a NETBLT implementation should be able to cope with
      out-of-sequence messages, perhaps collecting them for later
      processing, or even processing them immediately.  If an incoming
      control message "fills" a "hole" in a group of message sequence
      numbers, the implementation could even be clever enough to detect
      this and adjust its outgoing sequence value accordingly.

      When the control timer expires, the receiving NETBLT resends the
      control message and resets the timer.  After a predetermined
      number of resends, the receiving NETBLT can assume that the
      sending NETBLT has died, and can reset the connection.

      The sending NETBLT, upon receiving a control message, should act
      as quickly as possible on the packet; it either sets up a new
      buffer (upon receipt of an OK packet for a previous buffer),
      resends data (upon receipt of a RESEND packet), or sends data

      (upon receipt of a GO packet).  If the sending NETBLT is not in a
      position to send data, it sends a NULL-ACK packet, which contains
      a
      high-received-sequence-number as described above (this permits the
      receiving NETBLT to clear the control timers of any packets which
      are outstanding), and waits until it can send more data.  In all
      of these cases, the overhead for a response to the incoming
      control message should be small; the total time for a response to
      reach the receiving NETBLT should not be much more than the
      network round-trip transit time, plus a variance factor.

      The timer system can be summarized as follows: normally, the
      receiving NETBLT is working under one of two types of timers, a
      control timer or a data timer.  There is one data timer per buffer
      transmission and one control timer per control packet.  The data
      timer is active while its buffer is being transferred; a control
      timer is active while it is between buffer transfers.

      The above system still leaves a few problems.  If the sending
      NETBLT is not ready to send, it sends a single NULL-ACK packet to
      clear any outstanding control timers at the receiving end.  After
      this the receiver will wait.  The sending NETBLT could die and the
      receiver, with all its control timers cleared, would hang.  Also,
      the above system puts timers only on the receiving NETBLT.  The
      sending NETBLT has no timers; if the receiving NETBLT dies, the
      sending NETBLT will just hang waiting for control messages.

      The solution to the above two problems is the use of a death timer
      and a keepalive packet for both the sending and receiving NETBLTs.
      As soon as the connection is opened, each end sets a death timer;
      this timer is reset every time a packet is received.  When a
      NETBLT's death timer at one end expires, it can assume the other
      end has died and can close the connection.

      It is quite possible that the sending or receiving NETBLTs will
      have to wait for long periods of time while their respective
      clients get buffer space and load their buffers with data.  Since
      a NETBLT waiting for buffer space is in a perfectly valid state,
      the protocol must have some method for preventing the other end's
      death timer from expiring. The solution is to use a KEEPALIVE
      packet, which is sent repeatedly at fixed intervals when a NETBLT
      is waiting for buffer space.  Since the death timer is reset
      whenever a packet is received, it will never expire as long as the
      other end sends packets.

      The frequency with which KEEPALIVE packets are transmitted is
      computed as follows: At connection startup, each NETBLT chooses a

      death-timeout value and sends it to the other end in either the
      OPEN or the RESPONSE packet.  The other end takes the
      death-timeout value and uses it to compute a frequency with which
      to send KEEPALIVE packets.  The KEEPALIVE frequency should be high
      enough that several KEEPALIVE packets can be lost before the other
      end's death timer expires.

      Both ends must have some way of estimating the values of the death
      timers, the control timers, and the data timers.  The timer values
      obviously cannot be specified in a protocol document since they
      are very machine- and network-load-dependent.  Instead they must
      be computed on a per-connection basis.  The protocol has been
      designed to make such determination easy.

      The death timer value is relatively easy to estimate.  Since it is
      continually reset, it need not be based on the transfer size.
      Instead, it should be based at least in part on the type of
      application using NETBLT.  User applications should have smaller
      death timeout values to avoid forcing humans to wait long periods
      of time for a death timeout to occur.  Machine applications can
      have longer timeout values.

      The control timer must be more carefully estimated.  It can have
      as its initial value an arbitrary number; this number can be used
      to send the first control packet.  Subsequent control packets can
      have their timer values based on the network round-trip transit
      time (i.e.  the time between sending the control packet and
      receiving the acknowledgment of the corresponding sequence number)
      plus a variance factor.  The timer value should be continually
      updated, based on a smoothed average of collected round-trip
      transit times.

      The data timer is dependent not on the network round-trip transit
      time, but on the amount of time required to transfer a buffer of
      data. The time value can be computed from the burst rate and the
      number of bursts per buffer, plus a variance value <1>. During the
      RESENDing phase, the data timer value should be set according to
      the number of missing packets.

      The timers have been designed to permit reasonable estimation.  In
      particular, in other protocols, determination of round-trip delay
      has been a problem since the action performed by the other end on
      receipt of a particular packet can vary greatly depending on the
      packet type. In NETBLT, the action taken by the sender on receipt
      of a control message is by and large the same in all cases, making
      the round-trip delay relatively independent of the client.

      Timer value estimation is extremely important, especially in a
      high-performance protocol like NETBLT.  If the estimates are too
      low, the protocol makes many unneeded retransmissions, degrading
      performance.  A short control timer value causes the sending
      NETBLT to receive duplicate control messages (which it can reject,
      but which takes time).  A short data timer value causes the
      receiving NETBLT to send unnecessary RESEND packets.  This causes
      considerably greater performance degradation since the sending
      NETBLT does not merely throw away a duplicate packet, but instead
      has to send a number of DATA packets.  Because data timers are set
      on each buffer transfer instead of on each DATA packet transfer,
      we afford to use a small variance value without worrying about
      performance degradation.

   5.3. Closing the Connection

      There are three ways to close a connection: a connection close, a
      "quit", or an "abort".

      The connection close occurs after a successful data transfer.
      When the sending NETBLT has received an OK packet for the last
      buffer in the transfer, it sends a DONE packet <2>.  On receipt of
      the DONE packet, the receiving NETBLT can close its half of the
      connection.  The sending NETBLT dallies for a predetermined amount
      of time after sending the DONE packet.  This allows for the
      possibility of the DONE packet's having been lost.  If the DONE
      packet was lost, the receiving NETBLT will continue to send the
      final OK packet, which will cause the sending end to resend the
      DONE packet.  After the dally period expires, the sending NETBLT
      closes its half of the connection.

      During the transfer, one client may send a QUIT packet to the
      other if it thinks that the other client is malfunctioning.  Since
      the QUIT occurs at a client level, the QUIT transmission can only
      occur between buffer transmissions.  The NETBLT receiving the QUIT
      packet can take no action other than to immediately notify its
      client and transmit a QUITACK packet.  The QUIT sender must time
      out and retransmit until a QUITACK has been received or a
      predetermined number of resends have taken place.  The sender of
      the QUITACK dallies in the manner described above.

      An ABORT takes place when a NETBLT layer thinks that it or its
      opposite is malfunctioning.  Since the ABORT originates in the
      NETBLT layer, it can be sent at any time.  Since the ABORT implies
      that the NETBLT layer is malfunctioning, no transmit reliability
      is expected, and the sender can immediately close it connection.

6. MULTIPLE BUFFERING

   In order to increase performance, NETBLT has been designed in a
   manner that encourages a multiple buffering implementation.  Multiple
   buffering is a technique in which the sender and receiver allocate
   and transmit buffers in a manner that allows error recovery of
   previous buffers to be concurrent with transmission of current
   buffer.

   During the connection setup phase, one of the negotiated parameters
   is the number of concurrent buffers permitted during the transfer.
   The simplest transfer allows for a maximum of one buffer to be
   transmitted at a time; this is effectively a lock-step protocol and
   causes time to be wasted while the sending NETBLT receives permission
   to send a new buffer.  If there are more than one buffer available,
   transfer of the next buffer may start right after the current buffer
   finishes.  For example, assume buffer A and B are allowed to transfer
   concurrently, with A preceding B. As soon as A finishes transferring
   its data and is waiting for either an OK or a RESEND message, B can
   start sending immediately, keeping data flowing at a stable rate.  If
   A receives an OK, it is done; if it receives a RESEND, the missing
   packets specified in the RESEND message are retransmitted.  All
   packets flow out through a priority pipe, with the priority equal to
   the buffer number, and with the transfer rate specified by the burst
   size and burst rate.  Since buffer numbers increase monotonically,
   packets from an earlier buffer in the pipe will always precede those
   of the later ones.  One necessary change to the timing algorithm is
   that when the receiving NETBLT set data timer for a new buffer, the
   timer value should also take into consideration of the transfer time
   for all missing packets from the previous buffers.

   Having several buffers transmitting concurrently is actually not that
   much more complicated than transmitting a single buffer at a time.
   The key is to visualize each buffer as a finite state machine;
   several buffers are merely a group of finite state machines, each in
   one of several states.  The transfer process consists of moving
   buffers through various states until the entire transmission has
   completed.

   The state sequence of a send-receive buffer pair is as follows: the
   sending and receiving buffers are created independently.  The
   receiving NETBLT sends a GO message, putting its buffer in a
   "receiving" state, and sets its control timer; the sending NETBLT
   receives the GO message, putting its buffer into a "sending" state.
   The sending NETBLT sends data until the buffer has been transmitted.
   If the receiving NETBLT's data timer goes off before it received the
   last (LDATA) packet, or it receives the LDATA packet in the buffer

   and packets are missing, it sends a RESEND packet and moves the
   buffer into a "resending" state.  Once all DATA packets in the buffer
   and the LDATA packet have been received, the receiving NETBLT enters
   its buffer into a "received" state and sends an OK packet.  The
   sending NETBLT receives the OK packet and puts its buffer into a
   "sent" state.

7. PROTOCOL LAYERING STRUCTURE

   NETBLT is implemented directly on top of the Internet Protocol (IP).
   It has been assigned a temporary protocol number of 255.  This number
   will change as soon as the final protocol specification has been
   determined.

8. PACKET FORMATS

   NETBLT packets are divided into three categories, each of which share
   a common packet header.  First, there are those packets that travel
   only from sender to receiver; these contain the control message
   sequence numbers which the receiver uses for reliability.  These
   packets are the NULL-ACK, DATA, and LDATA packets.  Second, there is
   a packet that travels only from receiver to sender.  This is the
   CONTROL packet; each CONTROL packet can contain an arbitrary number
   of control messages (GO, OK, or RESEND), each with its own sequence
   number. Finally, there are those packets which either have special
   ways of insuring reliability, or are not reliably transmitted.  These
   are the QUIT, QUITACK, DONE, KEEPALIVE, and ABORT packets.  Of these,
   all save the DONE packet can be sent by both sending and receiving
   NETBLTs.

   Packet type numbers:

      OPEN:           0
      RESPONSE:       1
      KEEPALIVE:      2
      DONE:           3
      QUIT:           4
      QUITACK:        5
      ABORT:          6
      DATA:           7
      LDATA:          8
      NULL-ACK:       9
      CONTROL:        10

   Standard header:

      local port:       2 bytes
      foreign port:     2 bytes
      checksum:         2 bytes
      version number:   1 byte
      packet type:      1 byte
      packet length:    2 bytes

   OPEN and RESPONSE packets:

      connection unique ID:                   4 bytes
      standard buffer size:                   4 bytes
      transfer size:                          4 bytes
      DATA packet data segment size:          2 bytes
      burst size:                             2 bytes
      burst rate:                             2 bytes
      death timeout value in seconds:         2 bytes
      transfer mode (1 = SEND, 0 = RECEIVE):  1 byte
      maximum number of concurrent buffers:   1 byte
      checksum entire DATA packet / checksum
      DATA packet data only (1/0):         1 byte
      client-specific data:                   arbitrary

   DONE, QUITACK, KEEPALIVE:

      standard header only

   ABORT, QUIT:

      reason:       arbitrary bytes

   CONTROL packet format:

      CONTROL packets consist of a standard NETBLT header of type
      CONTROL, followed by an arbitrary number of control messages with
      the following formats:

      Control message numbers:

         GO:             0
         OK:             1
         RESEND:         2

         OK message:

            message type (OK):  1 byte
            buffer number:      4 bytes
            sequence number:    2 bytes
            new burst size:     2 bytes
            new burst interval: 2 bytes

         GO message:

            message type (GO):  1 byte
            buffer number:      4 bytes
            sequence number:    2 bytes

         RESEND message:

            message type (RESEND):     1 byte
            buffer number:             4 bytes
            sequence number:           2 bytes
            number of missing packets: 2 bytes
            packet numbers...:         n * 2 bytes

   DATA, LDATA packet formats:

      buffer number:                                4 bytes
      highest consecutive sequence number received: 2 bytes
      packet number within buffer:                  2 bytes
      data:                                         arbitrary bytes

   NULL-ACK packet format:

      highest consecutive sequence number received: 2 bytes
      acknowledged new burst size:                  2 bytes
      acknowledged new burst interval:              2 bytes

NOTES:

   <1>  When the buffer size is large, the variances in the round trip
        delays of many packets may cancel each other out; this means the
        variance value need not be very big.  This expectation can be
        verified in further testing.

   <2>  Since the receiving end may not know the transfer size in
        advance, it is possible that it may have allocated buffer space
        and sent GO messages for buffers beyond the actual last buffer
        sent by the sending end.  Care must be taken on the sending
        end's part to ignore these extra GO messages.