Completeness of regular inductive limits (Q908491)

After Enflo's negative solution to the approximation problem, Pisier's construction of an infinite dimensional Banach space \(P\) such that \(P\otimes_{\pi}P=P\otimes_{\epsilon}P\) and Taskinen's negative solution of the ``problème des topologies'' (and related questions on (DF)-spaces), the most important open problem arising from Grothendieck's work in the 1950's asks to clarify the exact relationship between regularity, quasi-completeness and completeness in (LF)-spaces. The following is one very interesting (and still unsolved) aspect of this more general problem: (*) If \((E_ n)_ n\) denotes an increasing sequence of Banach spaces such that each inclusion \(E_ n\to E_{n+1}\) is continuous and \(E=_{n}E_ n\) the inductive limit, must \(E\) regular (i.e., each bounded subset of \(E\) is contained and bounded in one of the spaces \(E_ n)\) already imply that \(E\) is complete? In many concrete cases, it is easy to show that an (LB)-space \(E\) (say, of spaces of sequences or functions) is regular while it turns out to be much harder to prove the completeness of \(E\). Hence it is to be expected that (*) has a negative solution, but so far all efforts to produce a counterexample have failed. In the article under review, the authors claim to give such an example of a regular (LB)-space \(E=_{n}E_ n\) which is not (quasi-) complete. They take \(E_ n\) to be the space of all \(x: \mathbb N\times\mathbb N\to \mathbb C\) such that \[ \| x\|_ n:=\max (\sup \{j^{-i}| x_{ij}|;\;i\leq n,\;j\in \mathbb N\},\;\sup \{| x_{ij}|;\;i>n,\;j\in \mathbb N\})<\infty \] and \(\lim_{j\to \infty}x_{ij}=0\) for \(i>n.\) This \(E\) is a sequence space and an inductive limit ``of Moscatelli type'' (in the sense of Bonet--S. Dierolf). Unfortunately, the authors' proof of the regularity of \(E\) is inconclusive (say, on p. 426, line -7 and p. 427, line 2), and in fact \(E\) is not regular, as has been pointed out by Bonet. Hence the desired example fails. Moreover, due to the work of J. Bonet and S. Dierolf on inductive limits of Moscatelli type, it is actually known that no counterexample to (*) can arise in this context; a completely different idea would be needed.