Free and projective generalized multinormed spaces (Q2084852)

The paper is not easy to follow. One reason is the mathematics involved, true. But the other is the awkward mix of presentation/notation/formulation the authors have chosen. Here are two of the main results in the paper: Theorem 3.5. Let \(M\) be an arbitrary set, and \(\mathbf{F}(M)\) the \(\odot\)-free \(\mathbf{L}\)-space with base \(M\) (cf. Theorem 2.17). Then: \begin{itemize} \item[(i)] \(\mathbf{F}(M)\) is \(\odot\)-projective and metrically projective. \item[(ii)] \(\mathbf{F}(M)\) is asymptotically \(\odot^0\)-projective and extremely projective. \end{itemize} And Theorem 3.7. \begin{itemize} \item[(i)] For every \(\mathbf{L}\) a retract, respectively near-retract of a well-composed \(\mathbf{L}\)-space is metrically, respectively extremely projective. \item[(ii)] If \(\mathbf{L}\) is properly presented relative to the family \(L^\nu\), then every metrically, respectively extremely projective \(\mathbf{L}\)-space is a retract, respectively near-retract of some well-composed \(\mathbf{L}\)-space. \item[(iii)] If \(\mathbf{L}\) is properly presented relative to the family \(L^\nu\), then every \(\mathbf{L}\)-space is a strictly coisometric image of a well-composed \(\mathbf{L}\)-space. \end{itemize} The paper is a whirlwind trip around the notion of projective objects in new Banach space categories introduced here, those of \(\mathbf{L}\)-spaces. This notion does not comes out of thin air; on the contrary, attempts to provide a sound ground for \(p\)-multinormed spaces, a kind of development of the operator spaces idea sown in [\textit{H. G. Dales} et al., Diss. Math. 524, 115 p. (2017; Zbl 1380.46011)]. An \(\mathbf{L}\)-space is obtained combining a vector space E, a normed space L and a contractive norm on the \(\mathfrak{L}(\mathbf{L})\)-module \(\mathbf{L}\otimes E\) (with outer product \(T\cdot(x\otimes e) = Tx\otimes e\)); namely, such that \(\|T\cdot u\|\leq \|T\| \|u\|\) for every \(T\in\mathfrak{L}(\mathbf{L})\) and \(u\in \mathbf{L}\otimes E\). The notion of \(\mathbf{L}\)-operator has to be defined accordingly. Now, transplanting the injective/projective notions from an algebraic context to a Banach spaces context is always problematic because the monic/epic notions are not well-suited; for instance, an epic arrow in Banach spaces is one having dense range and therefore it is impossible that even the one dimensional space K be projective: there is no way to lift an operator \(\mathbb{K}\to Z\) through a dense range operator \(X\to Z\). To get something one has to restrict oneself to work with surjective operators (and to thank the open mapping theorem) and therefore to work with an adapted form of projectivity. This is why, and understandable, the authors have to work only with their own kind of ``surjective \(\mathbf{L}\)-operators'' (the ``coisometric'' maps mentioned above), and adapted projectivity notions. Here it is where the ``\(\odot\)-projective, metrically projective, asymptotically \(\odot^0\)-projective, extremely projective\dots'' notions come from. However, neither those names nor the notation \(\square\), \(\bigcirc\) for functors, or \(\odot\) for a specific functor (see below), help. Then, a free object is a categorical notion, better connected in principle with adjointness than with projectivity. Once said that, and if one decides to take arms against that sea of troubles and by opposing end them\dots then the rewards start to appear. The authors begin to manage free objects as a kind of underdog projective objects, while Theorem 2.17 declares that if \(L^0_p(X)\) is the normed space of simple functions on a σ-finite measure space \(X\), and if \(\odot: L^0_p(X).spaces \to Set\) is a specific functor that attempts to capture the structure of a set \(M\) of \(L^\nu_p\)-spaces (\(\nu\in M\)), then \(\odot\) admits a free element if and only if the preassigned \(L^\nu_p\)-spaces are finite-dimensional. This free element will be later called \(\mathbf{F}(M)\), and we are now ready to read this report back from Theorem 3.5.