On the constant of Lipschitz approximability (Q6567154)

\textit{N. Ozawa} and the reviewer [Proc. Am. Math. Soc. 142, No.~5, 1681--1687 (2014; Zbl 1291.46013)]\Nformulated the question to know if every separable Banach space \(X\) contains a convex compact subset \(K\) which is generating, in the sense that its linear span is dense in \(X\), and moreover a Lipschitz-retract of the whole space \(X\). This problem happens to be strongly related with linear and non-linear approximation properties, as shown in particular by recent articles by \N\textit{P. Hájek} and \textit{R. Medina}, e.g., [Mediterr. J. Math. 20, No.~2, Paper No.~75, 13~p. (2023; Zbl 1511.46013)].\NThe work under review is a remarkable contribution to this field: The author shows that there exist an equivalent norm on \(l_1\) and some \(\lambda>1\) such that for this equivalent norm, no generating compact convex set is a Lipschitz retract of \(l_1\) with a Lipschitz constant less than \(\lambda\). The construction of this equivalent norm is done from scratch, and although the author tries his best to explain the proof, the arguments are difficult and deep. The space \(l_1\) equipped with Medina's norm has a basis constant larger than 1, and the construction points towards the existence of an equivalent norm on \(l_1\) which fails the metric approximation property. This would solve a problem which goes back to Grothendieck.