On Wolfgang Lusky's paper ``The Gurarij spaces are unique'' (Q6139353)

The aim of this paper is to survey Lusky's proof of uniqueness of the (separable) Gurariy space and discuss some related developments on the occasion of Archiv der Mathematik's 75th anniversary. Lusky published his proof in just this journal in 1976, see [\textit{W.~Lusky}, Arch. Math. 27, 627--635 (1976; Zbl 0338.46023)]. Gurariy proved in 1966 ([\textit{V.~I. Gurariĭ}, Sib. Mat. Zh. 7, 1002--1013 (1966; Zbl 0166.39303)]) the existence of a separable Banach space \(X\) with the following property: If \(E\) is any finite-dimensional subspace of \(X\) and you take some finite-dimensional superspace \(F\) of \(E\), then the isometry of \(E\) into \(X\) extends to an \(\varepsilon\)-isometry of \(F\) into \(X\), no matter how small \(\varepsilon>0\). In the same paper he also showed that any two such \(X\) are almost isometric. In Section~3 the author first sketches a proof that Gurariy spaces (separable or not) are Lindenstrauss spaces, i.e., \(L_1\)-preduals, using a characterization of such spaces from Lindenstrauss' famous 1964-memoir [\textit{J. Lindenstrauss}, Extension of compact operators. Providence, RI: American Mathematical Society (AMS) (1964; Zbl 0141.12001)]. Next, the representation of separable Lindenstrauss spaces as an increasing union of copies of \(\ell_\infty^n\) with admissible bases is presented. Section~4 is devoted to the main ideas of Lusky's proof based on the machinery from Section~3 (however omitting delicate technical details) as well as some interesting consequences. Theorem~4.4 is a nice example: Suppose \(x\) and \(y\) are two elements of the (dense in the unit sphere) set of smooth points of the unit ball of the Gurariy space. Then there is an isometric automorphism on \(G\) mapping \(x\) to \(y\). Another important fact is that the Gurariy space is universal among separable spaces. Poulsen proved the existence of a metrizable simplex for which its set of extreme points is dense in 1961, see [\textit{E.~T. Poulsen}, Ann. Inst. Fourier 11, 83--87 (1961; Zbl 0104.08402)]. In Section~5 we learn that the idea of the proof of the uniqueness of the Poulsen simplex by Lindenstrauss, Olsen and Sternfeld ([\textit{J.~Lindenstrauss} et al., Ann. Inst. Fourier 28, No.~1, 91--114 (1978; Zbl 0363.46006)]) comes from Lusky's proof. Other noteworthy ``old facts'' are: \begin{itemize} \item A space \(A(S)\) of affine continuous functions on a compact convex set \(S\) is a Lindenstrauss space if and only if \(S\) is a simplex. \item A separable Lindenstrauss space is the Gurariy space if and only if the extreme points of its dual unit ball forms a weak-star dense subset. \item If \(S\) is the Poulsen simplex, then \(A(S)\) is not the Gurariy space because the above mentioned Theorem~4.4 fails. However it can be somehow modified to become the Gurariy space. \end{itemize} A final Section~6 has some short ``outlook'' to Fraïssé theory and non-separable Gurariy spaces. The reviewer would like to mention yet another perspective on Gurariy spaces. As we have mentioned above, Gurariy spaces are Lindenstrauss spaces. \textit{H. Fakhoury}, in 1972 [J. Funct. Anal. 11, 436--452 (1972; Zbl 0252.46023)] characterized the Lindenstrauss spaces as those spaces that are locally 1-complemented in any superspace. In [\textit{T.~A. Abrahamsen} et al., Glasg. Math. J. 56, No.~2, 395--407 (2014; Zbl 1303.46012)], Theorem~4.3, it is shown that Gurariy spaces are exactly those Lindenstrauss spaces which are locally 1-complemented in any superspace by almost isometric local retractions. By Lusky's uniqueness result, the separable Gurariy space is the unique separable space with this property. It also follows that the Gurariy space has the Daugavet property.