wrote in message news:7136971a-4bc8-4f42-a605-5a37127ca4a9@googlegroups.com... Le jeudi 9 avril 2015 01:40:08 UTC+2, Paul Rubin a écrit : >> I'd say Ada's target market has shrunk in the last few decades, to the >> realtime systems and embedded control sectors. The remaining sector of >> big server-side non-realtime systems (a buddy of mine used to work on >> those) has been mostly ceded to Java, which is garbage collected. GC's >> main benefits are when the program is complicated enough that manual >> memory management increases development effort significantly and makes >> bugs more likely. It's less of an issue in simpler programs, especially >> those that don't need dynamic memory. >This seems to suggest there are only GC and manual memory management as >alternatives. Ada provides a third way in this respect with controlled >types. >Just my two centimes ;-) As I noted before, Ada provides at least 5 ways to manage dynamic memory: (1) Stack (2) Container (3) Controlled types (as in Smart Pointers) (4) Subpools (perhaps "semi-manual") (5) Traditional allocate/deallocate There's nothing "manual" about the first three from the perspective of a client (programmer). GC proponents complain about the work to create things like (1), (2), and (3) -- but there is no work for Ada programmers when you are using language capabilities or widely available libraries. Most people shouldn't be creating containers -- there's no point, you'll have a hard time doing better than the language-defined ones, and your time could be better used doing something else. Randy.