Efficient placement of audio data on optical disks for real-time applications

Reviewer: Ernst L. Leiss

Digital optical disks hold out a good deal of promise for data storage, although so far this promise remains largely unrealized. This paper motivates and formulates the problem of storing two relatively sparse data sequences as one denser sequence. A data sequence here consists of repeated data and no-data blocks (both of uniform fixed size); a sequence is sparse if the size of the no-data blocks is large compared with the size of the data blocks, and dense otherwise. (No-data blocks are required to introduce delays for time-dependent data such as audio; they can be, and are here, used for storing data from another sequence.) The paper then proceeds to give optimal ways of creating such a denser sequence, starting both from sequences of unlimited length and from sequences of given fixed length. These methods are quite simple and therefore have definite practical relevance. While the introduction is quite useful and clear, the remainder of the paper could become a classic of how not to write a paper. It contains far too much misleading jargon (for example, the notions of record, subrecord, pattern, word, and unit form a hierarchy), the results are buried under a mass of unnecessary, trivial combinatorial arguments (for example, Lemma 1 is totally unnecessary), and the obvious is extensively belabored (for example, the first 41 lines of the section on “Merging Records with Fixed Lengths” should be summarized in ten lines or less). In addition, Figures 2 and 3 are interchanged and the English could be improved. Most of this should have been taken care of by proper editorial control. In summary, I recommend starting with the conclusion (p. 870), then reading the first three pages (862–864) for motivation and notation, and skimming (or skipping) everything until the summary (cases 1, 2, and 3 on p. 870).