Mash-ups extract data fragments from disparate sources, and combine and transform the extracted fragments for display. Currently, mash-up developers tend to employ ad hoc representations for mash-up data, view mash-ups as applications, and use imperative scripts to extract and transform data fragments. These approaches can make mash-up development hard, and the mash-ups’ run-time performance poor. We address these concerns with our infrastructure to declaratively produce data mash-ups. In this paper, we introduce three parts of this infrastructure: Sixml, an XML language to uniformly represent a condensed mash-up; Sixml DOM, a means to manipulate and reconstitute mash-up parts on demand; and Sixml Navigator, an alternative path navigator to reconstitute and format a mash-up using queries in existing languages. We also present the highlights of an experimental evaluation.
Sudarshan Murthy, David Maier, Lois M. L. Delcambr