Originally posted by mrugiero
View Post
I will also add another thing: if let's say that the all low-level things are written into C (or C++), let's say 15.000 lines, and all the machinery is written around this low backend is written in a VM and is around 80.000 lines of AsmJS and JavaScript code generated from a Java/GWT backend, and these lines will generate C/C++ into 150.000 to 200.000 lines of code. Would be Mir a VM based server, or not? (think in the same terms about a kernel based logic, where the core of interrupts is written in C). (so: 15.000 lines of "core" and Java core logic -> 80K generated JS -> 200K lines of C++ that is linked within the kernel)
I think in true value of this phrasing is based on defining terms: "C++ is slow/complex/etc." which wording is not well defined by most cultures: "what slow means", "complex", etc. Is like "managed" in "VM world". As for me VM based runtimes as much they are using an ahead-of-time compilers (I will put GCJ as an example) gives very little differences of what "native" and "managed" would mean, both in performance terms, as how it runs and what it means. This is why I do think that using a language in which developers are comfortable with is the way to go for them, if they fill good enough the requirements of the tasks.
Comment