There is a rather intersting paper from Sorkin: http://arxiv.org/abs/hep-th/0504037 Ten Theses on Black Hole EntropyAuthors: Rafael D. Sorkin (Perimeter Institute and Syracuse University) (Submitted on 5 Apr 2005 (v1), last revised 19 Oct 2011 (this version, v2)) Abstract: I present a viewpoint on black hole thermodynamics according to which the entropy: derives from horizon "degrees of freedom"; is finite because the deep structure of spacetime is discrete; is "objective" thanks to the distinguished coarse graining provided by the horizon; and obeys the second law of thermodynamics precisely because the effective dynamics of the exterior region is not unitary. I just want to stress the following [Sorkin] Thesis 1: The most natural explanation of the area law is that S resides on the horizon. Thesis 4: The idea that the degrees of freedom are inside the black hole is wrong. Thesis 9: To understand the generalized second law requires a spacetime approach, not a canonical one. His reasoning regarding thesis 9 is striking and I would like to interpret as follows: 'locating' the degrees of freedom on a surface would require somehow to 'locate' this surface in spacetime; but for lightlike surfaces it is hopeless to 'locate' or 'embed' the surface in a spacelike slice or to reconstruct a non-spacelike horizon in a semiclassical approach on a spacelike slice which is used in canonical approach.