Abstract. Understanding the challenges faced by real projects in evolving variability models, is a prerequisite for providing adequate support for such undertakings. We study the evolution of a model describing features and configurations in a large product line--the Linux kernel variability model. We analyze this evolution quantitatively and qualitatively. Our primary finding is that the Linux kernel model appears to evolve surprisingly smoothly. In the analyzed period, the number of features had doubled, and still the structural complexity of the model remained roughly the same. Furthermore, we provide an in-depth look at the effect of the kernel's development methodologies on the evolution of its model. We also include evidence about edit operations applied in practice, evidence of challenges in maintaining large models, and a range of recommendations (and open problems) for builders of modeling tools.