Moving the kernel to modern C [LWN.net]

Moving the kernel to modern C

Posted Feb 27, 2022 0:09 UTC (Sun) by Wol (subscriber, #4433)
In reply to: Moving the kernel to modern C by camhusmj38
Parent article: Moving the kernel to modern C

> I would not say they are fools. But they are somewhat hidebound - viz the use of email as their primary collaboration system.

So you're another idiot who thinks newer = better.

I'm not saying these new-fangled things don't work for you. And plenty of kernel devs use newer tools. But all these idiots going "ooh! new! shiny!" make life hell for people doing the work.

I'm on a kernel mailing list. And reading the emails, I feel like tearing my hair out sometimes. But moving to Rust seems a far better solution than C++.

Cheers,
Wol

to post comments

Moving the kernel to modern C

Posted Feb 27, 2022 8:13 UTC (Sun) by camhusmj38 (subscriber, #99234) [Link] (2 responses)

Newer does not always equal better, but using technology to solve problems is literally the business we are in. Submission of patches and PRs by email has real implications - security for one, less open and effective collaboration for another. Again, I’m not saying that it wasn’t the right solution for its time but there are alternatives that work better.
And as for Rust, I’m a big fan of it but there is no way that the Kernel is going to be rewritten in Rust. It’s much more viable to replace some data structures and paradigms with C++ alternatives. This is possible without rewriting everything or completely retraining all existing contributors.
And I don’t think calling people idiots is particularly helpful to what is a technical discussion.

Moving the kernel to modern C

Posted Feb 27, 2022 10:37 UTC (Sun) by Wol (subscriber, #4433) [Link] (1 responses)

You just admitted yourself in another comment that "using C++ is not a technical problem". The problem is the social problem of making sure the C++ code really is smaller and more efficient than the C macros it replaces.

AND THERE ARE TOO MANY "OOH NEW SHINY" LEMMINGS...

Likewise moving away from email - the problem is a social problem - there AREN'T ENOUGH DEVELOPERS. I think a fair few subsystems ARE developed using solutions like github, gitlab, whatever. If it really worked, surely that model would spread rapidly. But it's not working, and it's not "not working" because the solution is better or worse, it's not working because there aren't enough people to make either solution work.

And actually, probably one of the biggest problems with C++, is that IT'S NOT TRANSPARENT. Developers don't have a clear model in their mind of HOW it works. I come from the days when tomorrow's weather forecast took a day to run on the most powerful computers if you were lucky! If you can't model performance down to the bare metal, you have no clue how long the program is going to take to run. That's one of the reasons the kernel has held back on compilers so long, the disconnect between engineering reality and theoretical correctness.

That's one of the reasons I rant about relational. With Pick the database is transparent - as an application developer I can REASON about performance right through to the OS. That's what's so hard with C++ - the devs can NOT reason through to the hardware. (Okay, it's getting harder even with C, but it's not obfuscated ...)

Cheers,
Wol

Stop this please

Posted Feb 27, 2022 16:51 UTC (Sun) by corbet (editor, #1) [Link]

Surely we can find a way to discuss things without calling each other idiots or lemmings, right? Please don't do this anymore.