The feature recognition problem is formulated as a geometric constraint satisfaction problem (GCSP), in which variables are the faces in the delta volume, the possible feature classifications determine values for the variables, and the constraints are the geometric and topological properties between the variables. A novel representation language is proposed to encode more, and somewhat redundant, information about geometric and topological properties in form features. The redundant information adds more constraints amongst objects in feature interactions and thus helps in feature recognition. Moreover, the language can be extended to represent new features.