Ellis on VSL
Point 1: Any VSL theory involving variable speed of photon travel must of necessity be based on some other method of measuring spatial distances than radar.
Point 2: Any VSL theory based on changes in the metric tensor components must explain how it differs from GR and how time and space measurements are related to the metric tensor.
Point 3: Any VSL theory involving a change in the limiting speed will not be Lorentz invariant; the way Lorentz invariance is broken must be made explicit.
Point 4: Any VSL theory involving a change in the speed of the photon travel must eventually propose some other equations than standard Maxwell's equations to govern electromagnetism.
Point 5: Any VSL theory must be done consistently in terms of its effects on the whole set of physical equations.
The critique takes a glaringly classical view of the cosmos, but one can hardly argue with the last four points. As far as popular VSL theories are concerned, I mostly agree with the critique. The first point, however, makes no sense at all! It is of course true that the very definition of the metre defines the speed of light locally, but this does not imply that we should compare measurements over cosmological distances using the same speed of light. How do we compare measurements over cosmological distances at all? The only work I am aware of which considers real measurements over large scales is Louise Riofrio's analysis of the speed of light on an older Earth. In fact, sensible approaches to quantum gravity take the measurement of distance via photon travel far more seriously than it is considered in classical gravity. Connes has expressed this nicely in his motivation for a new spectral physics. A varying speed of light is introduced by Riofrio as a useful picture for what we observe in the cosmos from here. Ellis can hardly argue with the observed expansion of space, or rather the observed increase in distance between stationary objects over cosmic time, which is based very much upon the spectral concept of distance.
Ellis mentions Penrose's approach to causality, so he probably won't argue with anybody who wants to replace Maxwell's equations immediately with sheaf cohomology in twistor theory. As Penrose himself says, twistor theory seems to have a bearing on quantum gravity, rather than general relativity itself. But it is well known that mass generation is a difficult question in twistor theory, although Hughston and Hurd made interesting progress in the late 1980s by combining two massless solutions to obtain an $H^2$ massive state. This work only stresses the importance of understanding higher dimensional non-Abelian cohomology, and in this framework a varying speed of light is really a minor concern.
The wording of Ellis's points betrays some further prejudices about the mathematics being used in VSL investigations. For example, the whole set of physical equations is far too restrictive a notion for a category theorist. Causality takes us beyond the realm of mere set theory, as we have seen.