Feature Algebra was introduced as an abstract framework for feature-oriented software development. One goal is to provide a common, clearly defined basis for the key ideas of feature-orientation. The algebra captures major aspects of feature-orientation, such as the hierarchical structure of features and feature composition. However, as we will show, it is not able to model aspects at the level of code, i.e., situations where code fragments of different features have to be merged. In other words, it does not reflect details of concrete implementations.In this paper we first present concrete models for the original axioms of Feature Algebra which represent the main concepts of feature-oriented programs. This shows that the abstract Feature Algebra can be interpreted in different ways. We then use these models to show that the axioms of Feature Algebra do not properly reflect all aspects of feature-orientation from the level of directory structures down to the level of actual code. This gives motivation to extend the abstract algebra, which is the second main contribution of the paper. We modify the axioms and introduce the concept of an Extended Feature Algebra. As third contribution, we introduce more operators to cover concepts like overriding in the abstract setting.Dedication It is our pleasure to dedicate this paper to José Nuno Oliveira on the occasion of his 60th birthday. José has been very active in the field of formal methods and, particularly, in algebraic techniques for that, such as relation algebra and category theory. Therefore we think it adequate to contribute a study on an algebra for feature-oriented system development, not only for cross-fertilisation, but hopefully also for enjoyment. Best wishes, José, for yourself, your work and for fruitful further interaction, as in the past years!