Don't care to have this degrade into "Quen es mas Phisico", but you yourself stated a set of assumptions regarding geometry and mass; hence it's an approximation.
While your statments as to the linear behaiviour of gravity is fact, it is still a fact based on approximation that the earth is uniformly dense and a perfect sphere. A true Integration on the Oblate Spheroid that is the Earth and most Planets would not give a uniformly linear increase in gravitational Force. Gravitational Mapping of the surface of the earth show clearly how it varies across the face of the earth.
Further, the 1/(r*r) behaiviour that we so carelessly apply is in Fact a point source approximation. We "Put" all the mass at the center of a planet/star as an approximation that works well when we are far away from the mass so that it "looks" like a point mass. Hence, all the planets can be considered points like in calculating their gravitational interaction and thus, their orbits and little-g.
However, if you are close to the mass, or as in this case inside, then this simple point source approximation fails and you acutally have to use, gulp, Calculus to integrate all of the point source around you, in essense every atom, to account for the mass around you to give you the gravity. The fact that we are considering a symmetrical uniformly dense object (to whit a sphere) means that this integration (which BTW is easiest in spherical coords
) gives us an easy result, the linear behaiviour you mentioned.
Maybe we just debating semantics here, but all I was addressing was that Newtons Universal Gravitation Law is for point sources and should be used with care on extended objects. Usually, nice clean answers like the behaiviour you described is based on one approximation or another....
Finally, to get a bit esoteric, there is mounting evidence that while Newtons Grav law works well at our distance scales, on the scale of the Quantum or the Cosmic, then it **seems** that the traditional f= (Gm1m2)/(r*r) actually fails; new facts ursurping old laws!