Linus Torvalds on C++ (2007) | Hacker News Reader

Linus Torvalds on C++ (2007)

harmful.cat-v.org

36 pointstosh22 comments

Different programming domains require different tools. Language wars stem from programmers incorrectly extrapolating the requirements of their domain to other domains that they're unfamiliar with. C is a good choice for kernel development. C is a bad choice for numerical computing. JavaScript is a good choice for web development. JavaScript is a bad choice for kernel development.

Some high profile C++ projects: LLVM, Webkit, v8 JavaScript engine, Chromium, Tensorflow, PyTorch, Eigen, OpenCV, Skia, Unreal Engine. C++ is a good fit for those projects: the right mix of expressiveness, portability & performance. It's not a good fit for other projects. Pick the right tool for the job. Move on with your life.

At least for me it's really fun reading Linus rant; but I wonder what his unkind ways mean for contributors to Linux.

I guess it's important to maintain a strong grip on the product when there are so many people touching the code, each one with different interests. But this kinds of attacks on (sometimes clueless) random would-be contributors just leaves me with a bad taste. And a community which is lead by someone who does this will only replicate that behavior.

I know I will probably never try and contribute anything to the kernel because of this toxicity; how many more people have taken this decision?

Strongly held beliefs, supported by well-reasoned arguments, are a positive signal for every project leadership. It shows that they care deeply and have a consistent vision about the direction. I'll take a passionate, driven and unfiltered position over pleasant mediocrity any day of the week.

Great projects (Python, Linux, OpenBSD, Apple) don't just have an idea what they want to be, but also what not to be. If you have no idea what not to be, you become C++ / Scala / Microsoft.

Strong tribes win.

At different phases of a project, you need different styles of leadership. While I think that was very funny at the time, 11 years later, I think with the maturing of linux, a very different style of leadership is needed than the brashness that won him fans over Microsoft over a decade ago.

How would a different type of leader improve the state of the Linux kernel? Honest question, I'm not that much into kernel development, just a happy consumer of their work.

I think it's great that they are focused and clear, but I'm specifically talking about hostility to random people.

This has been posted here several times before. No point discussing the same thing over and over again.

imho as time goes by ‘old’ posts gain new context & meaning

The last time it had discussion here was two years ago[1]. Perhaps its time for those who haven't seen it before to take a look?

[1] https://hn.algolia.com/?query=Linus%20Torvalds%20on%20C%2B%2...

I think what he says about C++ to a lesser or greater extent applies to most OOP languages, especially if speed is a concern. After a quarter century of OOP, including years with C++, I still do not understand why it exists or what problems it actually solves. I understand the problems it tries and pretends to solve but that's not the same thing. I still do not see the advantages (as opposed to "advantages") and the disadvantages of the approach are huge and glaring. I think the rise of OOP has led to a lot of programmers who don't know what they're doing because the languages themselves are created in such a way as to encourage bad practices like inheritance and make good practices like composition not obvious or intuitive.

Encapsulation. Hope this helps.

I'm not sure you know what you're talking about here.

If you want to see what OO-in-C might look like, look at golang's object model.

You lose some bits of OO design, but many folk just don't need/want them.

Exceptions, deep inheritance, etc.

Don't get me wrong, I'm well versed in C++/C#/Java, but you can quite easily paint yourself into a corner if you're not careful.

C programming requires much more careful thought, and that's probably why Linus isn't too keen on the OO crowd, as that's where, in his opinion, the sloppy programmers have wandered off to.

Why are you even talking about C let alone OO-in-C?

> After a quarter century of OOP, including years with C++, I still do not understand why it exists or what problems it actually solves.

This is what I was responding to.

Any decent GUI API will be OO. It is the right paradigm for that.

Abstract Syntax Trees for language translation is another problem where OO is a great choice.

Simulation is another good fit.

Wait, so does he like C++ or not?

Around linux 0.99 pl10 or 12, Linus wanted to compile the kernel using C++ compiler. I do not remember exactly the purpose. He has discovered many bugs in g++ and many slowness that were "impossible" to solve because of the existence of exceptions handling. g++ developpers were also not very cooperative.

He does not. See the other comments in this post.

I leave it here,

https://subsurface-divelog.org/

Torvalds explains in the original link that C++ is not appropriate for kernel modules. User space programs are another matter.

The core logic is in C, and that's not changing. The fact that QT is one of the few sane cross-platform UI choices did force some C++ code into there.

http://lists.subsurface-divelog.org/pipermail/subsurface/201...

I'll quote the relevant part here, since that's a long email:

  A word of warning: Linus has very strong feelings 
  about all the things that are wrong with C++ and
  at times has been known to be less diplomatic than 
  me when explaining his point of view... :-)

  But he made a clear statement that he is interested in 
  seeing this port happening, as long as most of the 
  program logic that is not UI code stays in 
  (quote) "sane C files". So please keep that in 
  mind as we drive this further.

If you read emails around that point of time, GTK+ development had really gone off the rails in everyone's viewpoint, and was frustrating Linus and everyone else.

Reply on news.ycombinator.com