Abstract

Andrew Myers: New abstractions for large-scale secure programming

New applications are large, distributed and concurrent, share complex structured data, and have demanding security and reliability requirements. But current systems abstractions do not give programmers the tools they need to write and reason about their code. A new programming model is needed that is simple and intuitive, yet exposes the underlying systems problems in a way that scales up to future applications. I will discuss three ongoing group projects exploring how programming language extensions can improve support for future application development:

- Secure program partitioning, in which the compiler automatically synthesizes distributed code in response to security annotations;

- Nested inheritance and intersection types, which support scalable code reuse for large systems;

- Diaspora, a language and system for sharing persistent objects across a large distributed system spanning trust domains.