> You are reading way more into his comment than I think was intended or implied.
You're likely correct (I'm unable to find the original context of the comment), but it's rarely employed with its original context. Indeed, I was more trying to say that I disliked the quotation itself (and how it can be used) than whatever Stroustrup meant by it.
> C++ is one of the most successful programming languages in history. There is a reason why.
And I said as much in my comment.
(I would also point out that some of the reasons may not be good reasons. But some certainly are!)
> Stroustrup has spent much of his career working to improve C++.
To improve the current situation, we may want to abandon C++ altogether for new programs rather than stubbornly hold on to it.
(In the back of my mind is Stroustrup's recent reaction to the NSA report on memory safety.)
I'm under no illusion that programmers will begin formally proving properties of their programs, as interested as I am in that. But people seem to rarely avail themselves of the tools that help them write correct C++, and rarely manage to use only modern C++, which suggests we may want to simply let go of it. Or radically change it – making safety features opt-out rather than opt-in.
I routinely deploy very large C++ applications used by international airlines around the world. And I haven't had a single bug in production for 5+ years. Not even one.
The secret is 9000+ auto tests pushing the code way beyond real world usage.
Changing from C++ to another language won't change the fact that those tests are needed. Memory leaks and multi-threading issues are simply non-issues for me. I use tools like Helgrind and Valgrind to find and fix those.
In other words, new languages like Rust doesn't solve any problems I have.
That's very impressive! I desperately wish more people did their production work like you do.
I genuinely can believe that Rust, Carbon, &c wouldn't make much of a difference for you. But for others who don't exercise the same discipline (for whatever reason, internal or external) I think they can. I agree there are no silver bullets; that's why each little bit helps.
edit: I think our software woes have mostly to do with human nature; solutions (if they exist) will primarily be educational and social, not technical. But universal computers are less than a century old, and we can certainly do better on both fronts. If we've learned something about the way most people write programs and interact with existing tools, we perhaps ought to make of use of that experience to design better languages and tools.
Nobody suggests you shouldn’t. Stroustrup has spent most of his career working on improving C++.
You are reading way more into his comment than I think was intended or implied.
C++ is one of the most successful programming languages in history. There is a reason why.