The field of AI is strewn with knowledge representation languages. The language designer typically designs that language with one particular application domain in mind; as subsequent types of applications are tried, what had originally been useful features are found to be undesirable limitations, and the language is overhauled or scrapped. One remedy to this bleak cycle might be to construct a representation language whose domain is the field of representational languages itself. Toward this end, we designed and implemented RLL-1, a frame-based Representation Language Language. The components of representation languages in general (such as slots and inheritance mechanisms) and of RLL-1 itself, in particular, are encoded declaratively as frames. By modifying these frames, the user can change the semantics of RLL-1's components, and significantly alter the overall character of the RLL-1 environment. Often a large Artificial Intelligence project begins by designing and implementing a high-level language in which to easily and precisely specify the nuances of the task. The language designer typically builds his Representation Language around the one particular highlighted application (such as molecular biology for Units (Stefik), or natural language understanding for KRL (Bobrow & Winograd) and OWL (Szolovits, et al.)). For this reason, his language is often inadequate for any subsequent applications, except those which can be cast in a form similar in structure to the initial task. What had originally been useful features are subsequently found to be undesirable limitations. Consider Units' explicit copying of inherited facts or KRL's sophisticated but slow matcher.