Extensor properties of orbit spaces and the problem of action extension (Q1358325)

In [\textit{S. A. Antonyan}, Math. USSR, Sb. 65, No. 2, 305-321 (1990); translation from Mat. Sb., Nov. Ser. 137(179), No. 3(11), 300-318 (1988; Zbl 0685.54013)] the following theorem was proved: Let \(G\) be a compact metric group, \(X\) be a \(G\)-equivariant absolute (neighbourhood) retract for all metric \(G\)-spaces. Then the orbit space \(X/G\) is an absolute (neighbourhood) retract for all metric spaces. In [\textit{S. A. Antonyan}, Russ. Math. Surv. 48, No. 6, 156-157 (1993); translation from Usp. Mat. Nauk 48, No. 6(294), 145-146 (1993; Zbl 0832.54019)] the result was extended to the case of arbitrary compact Hausdorff acting group \(G\). The main purpose of the paper under review is to provide an alternative proof to the above stated theorem for \(G\) an arbitrary compact Hausdorff group. The axis of the author's proof is Proposition 4 (an analogue of Theorem 4 in the first of the reviewer's papers cited above), which asserts the following: Let \(G\) be a compact Hausdorff group and \(X\) be a metric \(G\)-space. Then there exist a countable family of complete metric \(G\)-spaces \(C_i\), \(i=1,2, \dots\), and a metric absolute retract \(C\) such that (1) each \(C_i\) is \(G\)-equivariantly locally contractible and \(\dim (C_i/G) <\infty\), (2) \(X\) admits a closed \(G\)-equivariant embedding into the product \(C\times C_1\times C_2\times \dots\), endowed with the diagonal action of \(G\), where \(G\) acts trivially on \(C\). Unfortunately, the author's proof of Proposition 4 is not correct. Namely, the maps \(e_i:X\to C_i\), \(i=1,2, \dots\) are not well-defined (even for \(G\) the trivial group); the same set \(U_\nu\) can participate in two or more canonical triples \(\pi=(U_\mu, U_\nu, p)\in A_i\) with different maps \(\varphi_\pi: X\to L_\pi\), and this causes problems with the well-definition of the map \(e_i: X\to C_i\) following the author's rule: (1) \(e_i|_{U_\nu} =\varphi_\pi\) if \(\{\pi= (U_\mu,U_\nu,p)\), \(\varphi_\pi: X\to \text{Con} S_\pi \subseteq L_\pi\} \in A_i\), (2) \(e_i\) maps to the vertex of \(C_i\) all the other points of \(X\). The application of Proposition 4 makes also incorrect the proof of the main Theorem A and affects its corollaries. Besides, in the proof of Proposition 5 the finite-dimensionality of the orbit space \(D_i/G\) (a nontrivial fact) is not substantiated.