Because of the monotonic modular design of the Relax NG schemas, a schema for a sublanguage can be defined by not including some modules. If the LegalRuleML prefix module is not included, then the additional constraint regarding prefix evaluation is avoided. Note that CURIEs, and CURIE-like abbreviations, with empty prefix are still allowed - these should be treated as same-document relative references when converting to RDF. For this to work, The definition of a prefix mapping for the empty prefix should be disallowed. if the Consumer RuleML any_content module is not included, then there are no edges in the compact serialization that can carry the @index attribute. This avoids the additional constraint about sequential values of @index. When such instances are normalized, the correct value are given to @index attributes on the constructed edges. The other additional constraints appear to be unavoidable, at least with our current approach (e.g. different content models for @key on elements in the RuleML and LegalRuleML namespaces).