Data Communications Networking Devices - 4th Ed.pdf

4.3 INTELLIGENT MODEMS ________________________________________________________ 459assumptions about the data coding, and even learns that it is ASCII/English basedon the data itself. The compression tables are empty at the start of each connectionand are built as data is pased. When the data divides naturally into characters andthe `working set' of frequent characters is not too large natural language is goodhere), then adaptation will create a coding structure which compresses the datawell.Run-length encodingMultiple consecutive copies of the same character or 8-bit pattern) are compressedfor transmission using run-length encoding. If the encoder has sent the samecharacter three consecutive times, the encoder sends the count of the remainingidentical consecutive characters as a single 4-bit nibble. For example, a series of`A's would be sent as `AAA' with the remaining number of `A's sent as a 4-bitnibble, while, a series of ®ve `A's would be sent as three Huffman encodings of `A'and a nibble of binary two.DecodingTo decode an encoded data stream, the receiving modem assumes it has an exactcopy of the sending modem's compression data structures because:i) both modems reset their structures at the start of each connection;ii) communication has been error-free MNP assures this);iii) decoding the data stream causes identical data structure updates to matchthe encoder's.Since the receiving modem has the same compression tables as the sender, itknows which set of Huffman data decodes the incoming bit stream. The receivingmodem compares the incoming bits against the Huffman codes in the table untilthere is a complete match. At this time, the character is delivered to the DTE andthe table is updated.V.42 bisV.42 bis data compression is a modi®ed version of the Lempel-Ziv method ofcompression that was developed approximately 20 years prior to the promulgationof the recommendation. The V.42 bis compression method uses an algorithm inwhich a string of information received from an attached DTE is encoded as avariable length codeword. To facilitate the development of codewords strings arestored in dictionaries at the encoding and decoding device and are dynamicallyupdated to re¯ect changes in the composition of data.The key to the ef®ciency of the V.42 bis compression process is its dictionarywhich is dynamically built and modi®ed. The dictionary can be considered torepresent a set of trees in which each root corresponds to a character in thealphabet, as illustrated in Figure 4.56. Each tree represents a set of known stringsbeginning with one speci®c character, and each node or point in a tree represents