Building a XSLT Processor for large documents and high-performance.

XSLT is increasingly being used for processing very large XML documents. Existing implementation models are placing hard limits of the size of document that can be processed. We report on efforts to build a XSLT processor capable of handling Gigabyte sized documents with equivalent performance characteristics to the best known existing implementation models.

We show that by representing XML data using a record format with minimal inter-record links that the memory consumption of the XSLT processor can be significantly reduced without detrimentally effecting performance, and indeed it aids the processing performance of large documents. We also show that using a simultaneous XPath processing model over this format can further significantly increase performance for large documents while not hindering the performance of small document processing.