Extensors of noncompact Lie groups (Q1594147)

For each \(n\) the Banach-Mazur compactum \(Q(n)\) is the space of \(n\)-dimensional normed spaces (modulo the isometry relation) with the Banach-Mazur metric. Topologically, it is equivalent to the space of \(n\)-dimensional, centrally symmetric, convex bodies (the unit balls) (modulo \(GL(n)\)) with the Hausdorff metric. An old question is whether \(Q(n)\) is homeomorphic to the Hilbert cube. This question can be broken into two parts: (i) is \(Q(n)\) an absolute extensor? and (ii) does it have the \(m\)-dimensional disjoint disk property for every \(m\)? Recent work of several Russian mathematicians (the detailed history is given in the paper) has established that \(Q(n)\) is an absolute extensor. The proof extended known results on the action of compact Lie groups on a metric space to locally compact groups. Here the authors take a somewhat different approach using the Loewner-John ellipsoid and the orthogonal group to give another proof that \(Q(n)\) is an absolute extensor. They are also able to make some progress on the second (and hence the main) question: namely that there is an open dense subset of \(Q(n)\) (the equivalence classes of non-symmetric spaces) which is ``a variety modeled by the Hilbert cube''.