抄録
Recently, the sizes of XML documents have rapidly been increasing. Distributed XML is a novel form of XML document, in which an XML document is partitioned into fragments and managed separately in plural sites. Distributed XML documents can often be managed more easily than a single large document, according to geographical and/or administrative factors. In this paper, we consider performing XSLT transformation efficiently for distributed XML. To obtain an efficient method for this transformation, we devise an efficient evaluation method for XSLT pattern, which is a subset of XPath, and integrate it into top-down XSLT transformation. To evaluate an XSLT pattern for node v in a distributed environment, the site having v may need to access other sites many times. We propose two novel techniques to reduce such accesses; (1) precomputation of ancestors and (2) cache for predicate evaluation. We implemented our method in Ruby and made evaluation experiments. This result suggests that our method is more efficient than a centralized approach.
