Network Function Virtualization (NFV) aims to overcome the limitations of hardware middleboxes moving network functions to software. Ultimately, this approach enables deployment into virtualized environments with higher flexibility, reducing deployment time and costs. In common industrial practice, multiple smaller network functions are often chained together into service function chains. Yet, service function chains are extremely fragile, lacking basic guarantees on correct composition and dependency reasoning.

In this position paper, we propose a roadmap towards language abstractions that address this issue. We introduce the problems with current service function chaining technology and present the research program we envision.