Compressing XML with multiplexed hierarchical PPM models

We established a working Extensible Markup Language (XML) compression benchmark based on text compression, and found that bzip2 compresses XML best, albeit more slowly than gzip. Our experiments verified that T_XMILL speeds up and improves compression using gzip and bounded-context PPM by up to 15%, but found that it worsens the compression for bzip2 and PPM. We describe alternative approaches to XML compression that illustrate other tradeoffs between speed and effectiveness. We describe experiments using several text compressors and XMILL to compress a variety of XML documents. Using these as a benchmark, we describe our two main results: an online binary encoding for XML called Encoded SAX (ESAX) that compresses better and faster than existing methods; and an online, adaptive, XML-conscious encoding based on prediction by partial match (PPM) called multiplexed hierarchical modeling (MHM) that compresses up to 35 % better than any existing method but is fairly slow