Einstein is often touted as the physicist to annihilate the idea of the ether. This is peculiar, because it is as though the world stopped listening to his opinion on the matter prior to his reflections on general relativity (GR). Einstein never got too excited about proclaiming that an ether, after the conception of GR, is necessary; but he did, nonetheless, make clear arguments, the details, philosophical and historical, I will try to fill in —if only even a few of them.
To begin with, what was it originally that was problematic about the Maxwellian ether? In his Treatise on Electricity and Magnetism, James Clerk Maxwell made the commonsense inference that, if light (more generally, electromagnetic waves) are really waves, then they must be waving in something. Moreover, translational waves are not possible in fluids, so the “something,” the ether, must be a kind of solid, even if a jelly-like substance. A key component of the Treatise’s technical set up required an asymmetry in the phenomena of a magnet approaching a coil and a coil approaching a magnet. Einstein maintained that there is no phenomenal basis for claiming the existence of this asymmetry. Einstein was presented with a problem, leading up to his 1905 “On the Electrodynamics of Moving Bodies.” To maintain coordinate in variance and the constancy of the speed of light, he needed to do two things, one is very important to the present exposition, the other not so important at present, though it is, far and away, the historically and philosophically more important of the two. The objectively important resolution was to make the laws of physics absolutely invariant among all inertial reference frames by way of time dilation —the puzzle piece that was missing from Poincaré’s and H. A. Lorentz’ theories. The other thing that had to be done, which is the more important at present, is establishing a metaphysical basis for inertia: Newtwon’s third law required some point of traction, in the sense that a body in motion tending to remain in motion, within Newton’s system, would do so with respect to absolute space; and so, without absolute space, what would Einstein do to establish that such tendencies yet remain? He used Mach’s principle, the idea that stars and objects provide a trajectory preserving system of reference, though there is relative motion. The idea is that objects, particularly stars, if nothing else, provide a fixity that stands in the stead of absolute space, thus allowing for objects to vectorially tend to remain in motion. Absolute space is, as Einstein says around 1905 or thereabouts, “superfluous”.
Fast forward to 1920 and Einstein’s address at the University of Leyden. The casual reader of popular physics and the trained physicist, alike, will be shocked to read even a few lines of the given address. Einstein says such things as, “[t]hus the endeavor toward a unified view of the nature of forces leads to the hypothesis of an ether” and “[t]o deny the ether is ultimately to assume that empty space has no physical qualities whatever. The fundamental facts of mechanics do not harmonize with this view.” Einstein drives home his address with the earth-quaking statement which has, somehow, failed to reverberate throughout the physics community: “According to the general theory of relativity space without ether is unthinkable.” What changed between 1905 and 1915? The obvious answer, even stated explicitly in the quotes, is GR; but what about GR requires an ether? If we go back to the reasons for ether in early electrodynamics, we see that Maxwell and others saw a need for physical properties of space. What it was, originally, that encouraged Einstein to write voluminously against ether was the fact that the physical properties being assigned space were actually unnecessary predicates. “Immobility” and a “mechanical” nature —the predicates Maxwell, the British Maxwellians, and Poincaré, in particular, felt should be predicated to an underlying physical substance— were necessary for the furtherance of scientific study and the completeness of understanding of the physical world. It is precisely these that, even at Leyden, Einstein said were unnecessary. The adoption of Mach’s principle, and the acceptance of the notion that there is no preferred inertial reference frame, cemented that no immobile underlying substance was necessary. In this light, Bruce J. Hunt’s narrative, in The Maxwellians, is particularly revealing, because he catalogues, late in the book, how difficult of a time the physicists were having of making physical models that could accurately mechanically model the ether.
In actuality, there is a very marked difference between STR’s conception of space and GR’s, namely, that STR is, philosophically speaking, a completely relational space. Such a space was conceived of philosophers, like Leibniz, long ago. The brilliance of Mach, noting that there needs to be a context for vectorial inertial tendency, if one intends to replace Newton’s space, was embraced by Einstein; Leibnizian-styled relational, idealistic space supplemented by Mach’s principle satisfies all philosophical and physical criteria of available data and conceptual necessities. Such cannot be so for GR, because, as Einstein notes, terms must be predicated to space, within the framework of GR. Einstein’s speech on this point is not ardent enough, for the philosopher trained in GR will ask: When we predicate a word, such as “curvature,” a word only ever applied to material things, to space, what exactly does it mean? In today’s undergraduate physics lecture halls, the physics professors will say, “well, space is just nothing; so let’s get back to the physics.” However, even in the mathematics, predication arises —and Einstein noted this at Leyden, that the ten (Einstein) equations tell us the physical state of space. This is the difference between Leibnizian space (e.g., STR’s space) and GR’s space. When we have a space’s metric defined by:
it must be realized that “the structure of the space is not really determined until these functions gμν are really known.” This “gμν” determines the topological features of space in a way that further determine motion of objects. (Implicitly, we have a least action principle at work here.) In the simplest terms, varying on how Wheeler once said it, curved spaces tell objects how to move. Non-homogeneity of space implies predication of attributes.
Let’s put this in perspective, first logically, then historically. Einstein is saying that, like Maxwell’s thoughts, predicates are necessary for the appropriate physical understanding of space: space is physical, not “nothing”; it has states. The predicates have changed, but they seem more necessary fundamentally and to the mathematics than they ever were for early unification of light, electricity, and magnetism, all while these predicates are very different in nature. Instead of immobility and mechanical, Einstein proposes an ether that is not immobile or mechanical, but is curved and may have liquid-like properties (Lense-Thirring effects come to mind). Historically, what happened was certain predicates were thought necessary for understanding electrodynamics (Maxwell et al.), but these predicates were found unnecessary (STR); and then GR presented new predicates. It’s this last move that the physics community seems to have largely ignored, producing quite a few problems for the philosophically-minded graduate student of physics who calls a professor’s attention to the fact that predication of “curvature” to “nothing” is quite a silly idea.
Today, one does not encounter physicists talking about ether in the literature. Viewed as an historical error, the physics community — in the capacity of my living experience— does not like to revisit errors of the past. But this may be a case in which a revisiting is necessary, if only for the sake of fully exfoliating the meaning of a theory already in place. That’s the worst case scenario. The best case scenario is that it could stimulate physical inquiry in new directions.
 The important, assumed, piece of physics information, here, is that the modern physics teaches that magnetic field line flux induces current in a wire. This flux means that there is an invariance, in the sense that field lines cutting the wire, whether the coil is moving or the magnet is moving, will induce current. This is due to relative motion, whereas Maxwell and physicists of the period adhered to Newton’s absolute space.
 See either Nick Huggett’s chapter on Mach, with excerpts and commentary, in Space from Zeno to Einstein (p. 174-185); or see Mach’s The Science of Mechanics.
 All such quotations are taken from the 7 pages of the Leyden address in the first half of Sidelights on Relativity.
 The point of noting “states” of space, for the physics-un-initiated, is that space definitely must be homogenous and isotropic for space to remain a non-physical container.
 Einstein, Albert. “The Problem of Space, Ether, and the Field in Physics.” In Ideas and Opinions, Edited by Sonja Bargmann, 276-85. Translated by Sonja Bargmann. New York City: Three Rivers Press, 1982.
 Even H. A. Lorentz is moving in the direction of non-classical ether predicates, in his chapter “Ether” in Light Waves and Their Uses (1903). He proposes a wax-like substance, obviously trying to overcome the problem of translational waves needing a solid mechanical substance, and, in a way, foreshadowing post-GR predicates.
 There is also an issue that is picked up by a few scholars (physicists, philosophers, and historians), regarding the fact that Mach’s principle can’t fit into the project of general relativity, often noting Einstein’s conscious struggle to include it in the framework. See Max Jammer’s Concepts of Space. Therefore, I stress, in the above, the importance of Mach to STR’s conception of inertia, because, without the principle, inertia needs a contextual basis for its determination.