Extensional filters reveal algorithms

This contribution deals with the intensionality of programming languages, and especially addresses the question : "How to compare programming languages from an algorithmic point of view ?’’

We propose an indirect way to compare two languages computing a same set of functions, based on the use of a witness subset of programs, called a filter. If two languages compute the same set of function before filtering, but two different set after filtering, we argue that the initial languages were intensionally different and claim that the difference is in the algorithms they implement. Filtering thus provides a tool to perform an intensional observation of languages.