Section: Overall Objectives

Constructs for modular and isolating programming languages

Dynamically-typed programming languages such as JavaScript are getting new attention as illustrated by the large investment of Google in the development of the Chrome V8 JavaScript engine and the development of a new dynamic language DART. This new trend is correlated to the increased adoption of dynamic programming languages for web-application development, as illustrated by Ruby on Rails, PHP and JavaScript. With web applications, users expect applications to be always available and getting updated on the fly. This continuous evolution of application is a real challenge [40]. Hot software evolution often requires reflective behavior and features. For instance in CLOS and Smalltalk each class modification automatically migrates existing instances on the fly.

At the same time, there is a need for software isolation, i.e., applications should reliably run co-located with other applications in the same virtual machine with neither confidential information leaks nor vulnerabilities. Indeed, often for economical reasons, web servers run multiple applications on the same virtual machine. Users need confined applications. It is important that (1) an application does not access information of other applications running on the same virtual machine and (2) an application authorized to manipulate data cannot pass such authorization or information to other parts of the application that should not get access to it.

Static analysis tools have always been confronted to reflection [37]. Without a full treatment of reflection, static analysis tools are both incomplete and unsound. Incomplete because some parts of the program may not be included in the application call graph, and unsound because the static analysis does not take into account reflective features [46]. In reflective languages such as F-Script, Ruby, Python, Lua, JavaScript, Smalltalk and Java (to a certain extent), it is possible to nearly change any aspect of an application: change objects, change classes dynamically, migrate instances, and even load untrusted code.

Reflection and isolation concerns are a priori antagonistic, pulling language design in two opposite directions. Isolation, on the one hand, pulls towards more static elements and types (e.g., ownership types). Reflection, on the other hand, pulls towards fully dynamic behavior. This tension is what makes this a real challenge: As experts in reflective programming, dynamic languages and modular systems, we believe that by working on this important tension we can make a breakthrough and propose innovative solutions in resolving or mitigating this tension. With this endeavor, we believe that we are working on a key challenge that can have an impact on future programming languages. The language construct challenge tackled by RMoD is formulated as follows:

What are the language modularity constructs to support isolation?

In parallel we are continuing our research effort on traits (Traits are groups of methods that can be composed orthogonally to simple inheritance. Contrary to mixin, the class has the control of the composition and conflict management.) by assessing trait scalability and reuse on a large case study and developing a pure trait-based language. In addition, we dedicate efforts to remodularizing a meta-level architecture in the context of the design of an isolating dynamic language. Indeed at the extreme, modules and structural control of reflective features are the first steps towards flexible, dynamic, yet isolating, languages. As a result, we expect to demonstrate that having adequate composable units and scoping units will help the evolution and recomposition of an application.