When CIF was adopted as an information exchange standard by the IUCr in 1991, it was, if anything, ahead of the then state of the art. Free-format, extensible, low-overhead - it offered an easy route to implementation by scientific software, much still written in Fortran. By externalising the semantics of its tags to external dictionaries, it allowed ontology development to be decoupled from format, and thus had an important role to play in interoperability between purely crystallographic applications and the wider worlds of structural biology, chemical informatics and laboratory data management, as illustrated elsewhere in this symposium. The latest enhancements to CIF, with the introduction of an even more powerful dictionary definition language that support methods definitions and multiple data models, provide unrivalled potential for taking computer ontologies into completely new territories.