A Short History of Schema Mapping Systems
Giansalvatore Mecca, Paolo Papotti, Donatello Santoro · CINECA IRIS Institutional Research Information System (University of Basilicata) · 2012
There are many applications that need to exchange, correlate, and integrate heterogenous data sources. These information integration tasks have long been identified as important problems and unifying theoretical frameworks have been advocated by database researchers [5]. To solve these problems, a fundamental requirement is that of manipulating mappings among data sources. The application developer is typically given two schemas – one called the source schema, the other called the target schema – that can be based on different models, technologies, and rules. Mappings, also called schema mappings, are expressions that specify how an instance of the source repository should be translated into an instance of the target repository. In order to be useful in practical applications, they should have an executable implementation – for example, by means of SQL queries or XQuery scripts. This latter feature is a key requirement in order to embed the execution of the mappings in more complex application scenarios, that is, in order to make mappings a plug and play component of integration systems. Traditionally, data transformation has been approached as a manual task requiring experts to understand the design of the schemas and write scripts to translate data. As this work is time-consuming and prone to human errors, mapping generation tools have been created to make the process more abstract and user-friendly, thus easier to handle for a larger class of people. In this paper, we outline a history of the different phases that have characterized the research about automatic tools and techniques for schema mappings and data exchange. We identify three different ages, as follows.