Paradoxically, as storage capacity and transmission bandwidth become cheaper and more readily available, interest in methods of data compression seems to grow. Indeed, the subject has never been more popular, as evidenced by the publication of several recent books and survey papers and by the unexpectedly large attendance at the first Conference on Data Compression in 1991.
This volume is the third edition of one of the earliest books on the subject (first published in 1984 [1]), and the fact that it has proven worthwhile to publish a new edition attests to the success of its predecessors. It is a practitioner’s book and includes numerous programs in BASIC (they can be obtained from the publisher on a diskette). The programs are the principal addition to the first edition of the book. Most of them are simple; some of them address problems that are trivial. Nonetheless, they may interest those without much programming experience who wish to experiment with practical data compression.
Most of the compression techniques covered are ad hoc and old. The reference list comprises only 11 items, the most recent being from 1978; although a much more extensive bibliography is appended that includes a handful of items from the 1980s, these are not cited in the text. An early chapter on data compression techniques describes simple ideas such as null suppression, bit-mapping, run-length coding, half-byte packing, a limited form of digram (here called “diatomic”) coding, replacement of common words by short codes, and difference coding (for numeric data).
The next chapter, “Statistical Encoding,” is an extensive presentation of Huffman coding. An unfortunate addition to the first edition of the book is an account of the universally inferior Shannon-Fano code, which was made obsolete in the early 1950s. In a confusing section (pp.160–162) a code for a particular set of symbol frequencies is exhibited and called a “Huffman code.” This code performs worse than the Shannon-Fano code for the same data and is not a Huffman code after all. A brief section on “adaptive compression” describes adaptive Huffman coding.
The strongest aspect of this book is its practical orientation. A welcome new section entitled “Modern Statistical Compression Methods” describes two compression protocols: the CCITT V.42 bis recommendation, which uses the LZ (Ziv-Lempel) coding method (although this is not a statistical compression technique) and the Microcom Networking Protocol (not being familiar with Microcom Inc., I would have welcomed some information on where the protocol is being used). Held also describes a program that reports on the susceptibility of particular files to compression by different methods, but unfortunately only trivially simple methods that are not likely to be used in practice are included.
A section on hardware products for compression gives interesting accounts of two commercially available stand-alone compression units and two statistical multiplexers that incorporate compression. Some software products are also described, including the ARC file compression utility commonly used in personal computers (but not the widely-used UNIX compress program). A large amount of the information about ARC appears to have been reproduced directly from the user manual.
In summary, this practically oriented book on text compression may serve to complement some of the recently published works that emphasize principles. Its coverage of modern compression techniques is spotty, however--for example, arithmetic coding, statistical context modeling, and the numerous LZ variants are not even mentioned--and this will inevitably limit its value.
