The goal of CML is to provide a common syntax with a well-defined semantics so that different implementations, with different internal representations and inference procedures, will be able to accurately parse the same domain theory. Ideally, inferences based on the same scenario and domain theory, but made by different implementations, would be consistent. Unfortunately, such consistency is contingent on a number of assumptions that are currently implicit in the various implementations. These caveats include modeling assumptions, relevancy assumptions, order of magnitude assumptions, the completeness of the specific inference procedures employed, and the kind of closed world assumption (CWA) that is made. In order to draw useful inferences, implementations must make these sorts of assumptions. The work of precisely characterizing the differing assumptions and mapping between them, however, has only just begun. This makes it difficult to directly compare inferences drawn by different implementations. Nonetheless, we expect that the inferences made by various implementations will be sensible (given the domain theory) and comparing them will help substantially in identifying the differences and commonalities among them.
The syntax and semantics of CML top level forms is described in
section 
.