Generalizing Parametricity Using Information Flow (Extended Version)Technical Reports (CIS)
Document TypeTechnical Report
Date of this Version6-24-2005
AbstractRun-time type analysis allows programmers to easily and concisely define operations based upon type structure, such as serialization, iterators, and structural equality. However, when types can be inspected at run time, nothing is secret. A module writer cannot use type abstraction to hide implementation details from clients: clients can determine the structure of these supposedly "abstract" data types. Furthermore, access control mechanisms do not help isolate the implementation of abstract datatypes from their clients. Buggy or malicious authorized modules may leak type information to unauthorized clients, so module implementors cannot reliably tell which parts of a program rely on their type definitions. Currently, module implementors rely on parametric polymorphism to provide integrity and confidentiality guarantees about their abstract datatypes. However, standard parametricity does not hold for languages with run-time type analysis; this paper shows how to generalize parametricity so that it does. The key is to augment the type system with annotations about information-flow. Implementors can then easily see which parts of a program depend on the chosen implementation by tracking the flow of dynamic type information.
Citation InformationGeoffrey Washburn and Stephanie C Weirich. "Generalizing Parametricity Using Information Flow (Extended Version)" (2005)
Available at: http://works.bepress.com/stephanie_weirich/1/