The expressiveness of a programming language is informally characterised as the ease with which ideas and concepts are naturally mapped into the language.

This gap between intuition and formalisation can be reduced with good languages but clearly not completely removed. However, we argue that even when we cannot express the concepts themselves it is nevertheless useful to express their existence and relationships with other concepts.

I will introduce extremely simple models, inspired by ontology models and refinment models, but making no assumption whatsoever about the internal structure (if any) of the involved abstraction levels. These models describe the informal part of a program in a way that can be used for actual computation and software composition. They can naturally capture computations crossing abstraction levels (compilation and interpretation, debugging and decompilation, etc.). They are centered around the notion of identity -- e.g. in which sense can we say that an image encoded in JPEG is "the same" as the original one given that JPEG is a lossy file format? Finally, we will see how they may lift the need for standardisation, including (reflectively) for their own formalism.

This seminar can be heard as a case against standards in the computer research and industry, defending the point of view that they are unnecessary and counter-productive for software composition, interoperability, and data exchange formats and protocols.