Processing Continuous Historical Queries over XML Update Streams

We address the problem of processing continuous historical queries over streams of XML data, returning continuous, exact answers. The stream data considered are tokenized XML data with embedded updates for inserting, removing, or replacing stream subsequences that correspond to complete XML tree nodes when they are fully materialized to trees. Our query language for expressing historical queries is XQuery extended with temporal annotations that specify sliding windows that focus on the stream results of query subexpressions, rather than on the query input streams. We have developed a novel architecture that can evaluate historical queries over large XML data streams using short history lists. Instead of buffering a fixed-size window, sliding through the input stream events, we focus on the temporal requirements of individual operations derived from the query and use history lists to cache the state history of the operation as long as necessary to answer the query.