There would not have been an economics of convention (EC) without the use of the word "convention" in chapter 12 of the "The General Theory of Employment, Interest, and Money" (1936) by Keynes, and without the book "Convention. A Philosophical Study" (1969), by the philosopher and mathematician David Lewis. But representatives of EC reinterpret the usual reading of those two texts. They extract from the first one the idea of a convention as regulating a professional community (the financial one and the academic one in economics). As for the second one, they privilege the final revision of Lewis' initial game-theoretic definition, which puts non-observable "beliefs" on a par with observable "actions." The coherence between both elements can only be produced by the emergence of a "(social) practice." Therefore a very different practice of economics is promoted by EC (for instance reunifying coordination and reproduction). Following Foucault who studied states as a practice (through the notion of "governmentality"), we study business firms as a practice. Because of the gap between the legal person (corporation whose members are the share-holders) and the economic organization (with all its stake-holders), the firm as a practice needs to be regulated by a convention, in order to make the inequality not unbearable for workers. Otherwise the working of the firm as a dispositive of collective creation would be blocked. We conclude that conventions, practices, and dispositives belong to the same analytical space.