> In RAII aware languages (it is not only C++ that has it), you don't need to forget to call anything.
Uhm, how about delete? C++ only introduced working smart pointers a quarter century in with C++11.
In the case that the lifetime coincides with the lexical scope C++ is robust, but so is, IMO, the with-* macro approach even without static validation (because unlike Go-style defer the construct itself introduces the object you want, so there is nothing to forget; if you really only care about the lexically scoped case you can also make this the only public API, and rule out errors that way).
I think the most compelling argument that can be made for RAII is that GC is a bad abstraction, because it special cases memory deallocation whereas "true" RAII can deal with fairly arbitrary resources (with the big caveat that de-allocation must never fail), but to really make a strong form of this argument I think you need something like Rust's ownership model; even the weak form needs more than C++<11 had.
> It definitely was, hence why everyone was replacing C with C++ (Apple, Microsoft, IBM, Be) until the FOSS freedom fighters decided to push C as their main language.
Beauty had nothing to do with it. Would I write serious software in C over C++? No, just for less terrible strings and resource management alone. But would you have picked C++ in the early 90ies over Turbo Pascal because of its greater beauty, or even substantial technical reasons?
The case of the enlightened companies vs unwashed FOSS masses is also a bit more complicated, and it's not that obvious to me the FOSS people made the wrong choice.
Microsoft: their software at the time was of notoriously low quality; early linux was significantly more stable (and performant). Be: I think I recall an ex-employee claiming on HN that their OS (despite innovative ideas) also had major quality issues. Apple: they sure did a lot of work over the years to avoid having to commit to C++. IBM: what software did they even write during that time? OS/2?
There are also major interfacing issues to other languages. And massive conceptual complexity, and difficulty agreeing on some "sane subset" over a huge community. Finally modern C++ these days takes, what, 100x longer to compile than C? I think a big part of the success of open source is due to the development of new languages with vibrant eco-systems, I suspect a massive move to C++ would have made that much more difficult.
Uhm, how about delete? C++ only introduced working smart pointers a quarter century in with C++11.
In the case that the lifetime coincides with the lexical scope C++ is robust, but so is, IMO, the with-* macro approach even without static validation (because unlike Go-style defer the construct itself introduces the object you want, so there is nothing to forget; if you really only care about the lexically scoped case you can also make this the only public API, and rule out errors that way).
I think the most compelling argument that can be made for RAII is that GC is a bad abstraction, because it special cases memory deallocation whereas "true" RAII can deal with fairly arbitrary resources (with the big caveat that de-allocation must never fail), but to really make a strong form of this argument I think you need something like Rust's ownership model; even the weak form needs more than C++<11 had.
> It definitely was, hence why everyone was replacing C with C++ (Apple, Microsoft, IBM, Be) until the FOSS freedom fighters decided to push C as their main language.
Beauty had nothing to do with it. Would I write serious software in C over C++? No, just for less terrible strings and resource management alone. But would you have picked C++ in the early 90ies over Turbo Pascal because of its greater beauty, or even substantial technical reasons?
The case of the enlightened companies vs unwashed FOSS masses is also a bit more complicated, and it's not that obvious to me the FOSS people made the wrong choice.
Microsoft: their software at the time was of notoriously low quality; early linux was significantly more stable (and performant). Be: I think I recall an ex-employee claiming on HN that their OS (despite innovative ideas) also had major quality issues. Apple: they sure did a lot of work over the years to avoid having to commit to C++. IBM: what software did they even write during that time? OS/2?
There are also major interfacing issues to other languages. And massive conceptual complexity, and difficulty agreeing on some "sane subset" over a huge community. Finally modern C++ these days takes, what, 100x longer to compile than C? I think a big part of the success of open source is due to the development of new languages with vibrant eco-systems, I suspect a massive move to C++ would have made that much more difficult.