http://www.w3.org/2003/08/binary-interchange-workshop/31-oracle-BinaryXML_pos.htm XMLZip XMLZip is a compressor and decompressor for XML documents based on the W3C DOM. XMLZip is written in Java and produces ordinary pkzip/WinZip zip files [7]. XMLZip first parses XML data with a DOM parser, then breaks the structural tree into multiple components: a root component containing all data up to depth N from the root, and one component for each of the subtrees starting at depth N. The root component is then modified, references to each subtree are added onto the root, and finally components are compressed.
XMLZip allows users to choose the depth at compression time, thus allowing users to select the DOM level at which to compress the XML files. This allows continued use of the DOM API without decreased performance. XMLZip only decompresses the portion of the XML tree that needs to be accessed, allowing applications to access data without uncompressing the whole document, reducing execution time and run-time space and memory usage.
XMLZip reduces the size of XML file while keeps the accessibility of the DOM API. However, XMLZip can only be run on entire XML file objects, and is thus offline-only. Also, XMLZip compression ratio is not as good as gzip when measured over an entire document.
Было здесь, но видать куда-то уже делось
http://www.xmls.com/resources/xmlzip.xml Открыть - наверное имеет смысл искать запакованные блоки (может есть сигнатуры PK...), а потом открывать каждый обычным zip-алгоритмом.
