The OpenCPU System: Towards a Universal Interface for Scientific Computing through Separation of Concerns
This addresses interoperability issues in scientific computing for researchers and developers, though it appears incremental as it builds on existing technologies like HTTP and R.
The paper tackles the problem of integrating analysis components across applications by proposing a universal interface for scientific computing, resulting in the OpenCPU system which uses HTTP and R to provide an abstracted API.
Applications integrating analysis components require a programmable interface which defines statistical operations independently of any programming language. By separating concerns of scientific computing from application and implementation details we can derive an interoperable API for data analysis. But what exactly are the concerns of scientific computing? To answer this question, the paper starts with an exploration of the purpose, problems, characteristics, struggles, culture, and community of this unique branch of computing. By mapping out the domain logic, we try to unveil the fundamental principles and concepts behind statistical software. Along the way we highlight important problems and bottlenecks that need to be addressed by the system in order to facilitate reliable and scalable analysis units. Finally, the OpenCPU software is introduced as an example implementation that builds on HTTP and R to expose a simple, abstracted interface for scientific computing.