Classical thermodynamics, including a basic concept of entropy satisfying dS = dQ/T, where dQ is not an exact differential but dS is exact, is a well developed theory not involving a concept of information or of statistics. It is enormously useful in physics, chemistry and many branches of engineering. It isn't much used in computer science, so most computer scientists manage to remain ignorant of it. Classical thermodynamics won't go away any more than Newtonian mechanics will go away. [Remark: While entropy applies to equilibrium states, the concept has been stretched to usefully fit some rather rapid processes, e.g. in aerodynamics where equilibria changing over microseconds are useful concepts.] Already in the middle of the 19th century, Maxwell and Boltzmann undertook to base thermodynamics on statistical mechanics and kinetic theory of gases. They did not succeed in getting everything they wanted about classical thermodynamics from statistical mechanics, and their relations remain a subject of research and controversy even today. None of this is in opposition to most of the postings on the subject, and I rather like the notion entropy being connected with initial conditions.