›24;H
            <<< EISNER::DUA16:[NOTES$LIBRARY]HARDWARE_HELP.NOTE;1 >>>
                               -< HARDWARE_HELP >-
================================================================================
Note 674.8       What does MNP and error correction really mean?          8 of 8
EISNER::TILLMAN "DECUServe MoS Lame Duck"            57 lines   3-DEC-1990 16:48
                           -< MNP Class Definitions >-
--------------------------------------------------------------------------------
    I received a document from Microcom today explaining what each class of
    MNP is supposed to do.  Here are some excerpts:
    
    Class 1 - This is the first level of MNP performance, using an
    asynchronous byte-oriented half-duplex method of exchanging data.  This
    class makes it possible for devices with few hardware resources to
    communicate error-free.  This class is no longer implemented in Microcom
    modems.
    
    Class 2 - This class uses asynchronous byte-oriented full-duplex data
    exchange.
    
    Class 3 - This class uses synchronous bit-oriented full-duplex data
    exchange, eliminating start and stop bits.  The modem communicated with
    the DTE asynchronously and with the other modem synchronously.
    
    Class 4 - This class uses Adaptive Packet Assembly (TM) which alters the
    size of the packets being sent depending on line quality and Data Phase
    Optimization (TM) which removes repetitive control information from the
    data stream.  When combined with Class 3, this will yield about 120%
    efficiency.
    
    Class 5 - This class implements basic data compression, using a
    real-time adaptive algorithm to do the compression.  A net throughput of
    200% can be realized with this class.
    
    Class 6 - This class implements Universal Link Negotiation (TM) to
    choose the proper data speed and Statistical Duplexing (TM) to simulate
    full-duplex service on a half-duplex V.29 connection.
    
    Class 7 - This class implements a different compression scheme than is
    used in Class 5, allowing a 300% throughput compared to a non-MNP modem
    connection.  The compression scheme is Huffman encoding with a
    predictive algorithm.  In addition to the Class 5 and 7 compression,
    V.42bis compression is also supported.  This compression is based on the
    Limpel-Ziv algorithm and can produce a 400% throughput increase.
    
    Class 9 - This class reduces the time required to acknowledge a received
    message by piggy-backing an acknowledgement on a following data packet,
    if there is one, and to retransmit a message following an error by
    indicating in the NAK the packet number of the failed packet.  This
    causes the retransmitting modem to resend only the packet that failed.
    
    Class 10 - This class includes Adverse Channel Enhancements (TM) that
    optimize performance on lines with varying quality, such as cellular
    phones, international lines, or rural lines.  The enhancements include:
    
    o	Multiple aggressive attempts at link setup
    o	Adapting packet size to varying levels of interference
    o	Negotiating transmission speed shifts to achieve maximum acceptable
    	line speed
    o	Dymanic shifting of modem speed for line conditions
    
    Classes 1 through 4 are in the public domain and a spec is available for
    $100 from Microcom, Inc.