When employing custom attributes to store meta-data, is it best to decorate the interface, or the class that implements the interface, assuming that any class that implements the interface would have the same data in the attribute?
Update: Basically i'm writing a custom data storage mechanism for a project, and the objects represent the various tables being stored. The custom attribute is used to designate which table in the dataset is used to store the objects of that class, and also to identify which tables are involved in a n:m relationship.
So if i put the attributes on the interface, is this a clearer approach, or does it clutter the interface and make accessing the data itself more cumbersome?