Large-Scale Microtask Programming
This addresses the challenge of coordinating crowd workers in software development, particularly for microservices, but it appears incremental as it builds on existing microtasking concepts.
The authors tackled the problem of making microtask programming more efficient and reducing conflicts by introducing a behavior-driven approach where developers identify, test, implement, and debug behaviors in isolation, and they developed the first method for building microservices through microtasks, implementing it in a web-based IDE called Crowd Microservices.
To make microtask programming more efficient and reduce the potential for conflicts between contributors, I developed a new behavior-driven approach to microtasking programming. In our approach, each microtask asks developers to identify a behavior behavior from a high-level description of a function, implement a unit test for it, implement the behavior, and debug it. It enables developers to work on functions in isolation through high-level function descriptions and stubs. In addition, I developed the first approach for building microservices through microtasks. Building microservices through microtasks is a good match because our approach requires a client to first specify the functionality the crowd will create through an API. This API can then take the form of a microservice description. A traditional project may ask a crowd to implement a new microservice by simply describing the desired behavior in a API and recruiting a crowd. We implemented our approach in a web-based IDE, \textit{Crowd Microservices}. It includes an editor for clients to describe the system requirements through endpoint descriptions as well as a web-based programming environment where crowd workers can identify, test, implement, and debug behaviors. The system automatically creates, manages, assigns microtasks. After the crowd finishes, the system automatically deploys the microservice to a hosting site.