Rust Linux Developers Compared To Road Builders & Mapmakers

[phoronix]

Phoronix: Rust Linux Developers Compared To Road Builders & Mapmakers

Longtime Direct Rendering Manager (DRM) subsystem maintainer David Airlie of Red Hat has written an interesting blog post providing an analogy to types of developers compared to road builders and hotels...

https://www.phoronix.com/news/Airlie-Rust-Linux-Dev-Compare

[dragorth]

So, just reading Michael's quote of the blog so far, I would say there is a fundamental issue of the hotel builders saying the roads are already built, the tooling they use already works for them, they are able to get to their hotels, never mind they rode a donkey and the Kernel Devs need to bring in cranes and other tools for the much larger hotels, theme parks, high rises and other things they still have to build, whether rust exists or not.

I am now going to read the blog post to see if any of this is covered.

[User29]

I'm always happy to read Asahi devs opinion on software development because they seem to be competent, experienced and have already delivered stuff.

[AHSauge]

So basically, some people cling to their (almost) 50 year old dinosaur like their lives depend on it. I get that new tech is scary for some, and I wholeheartedly agree Rust (in this case) isn't a magic fits-everywhere type of solution, but jeeeez, "This one driver works, so if you make it exactly like that, yours will work too" is seriously not how you make robust and reliable software. It's a careless way of working where you basically don't want to attend to anyone else's needs. That's just not cool at all.

[Alexmitter]

It was just merged way too early. It shows the entitlement and "our way or the highway" mentality in the rust fanbase.
Writes a driver their way, gets told thats not how we do that here, goes crying on mastodon. Either you do it the established way or you properly explain why you want to do be the one special driver.

I believe it is best to remove the rust support for now, until at least there is a consortium based specification for the language and a good compiler implementation. It is just fully unacceptable that I now need GCC and LLVM just so I have one compiler being good at C and one compiler because I have no other choice.

[kpedersen]

s/roads/bindings/g

When I wrote the DRM scheduler abstractions, I ran into many memory safety issues caused by bad design of the underlying C code. The lifetime requirements were undocumented and boiled down to "design your driver like amdgpu to make it work, or else".

s/abstractions/bindings/

And those memory safety issues in the C code, are still there... Rust now just sits ontop and no safety was gained. The Rust code now enforces "fantasy" lifetime guarantees that can still effectively dangle and cause use-after-free errors when the underlying C data gets stripped.

[oleid]

And those memory safety issues in the C code, are still there...

Do you know this or are you speculating?

[V1tol]

So basically, some people cling to their (almost) 50 year old dinosaur like their lives depend on it. I get that new tech is scary for some, and I wholeheartedly agree Rust (in this case) isn't a magic fits-everywhere type of solution, but jeeeez, "This one driver works, so if you make it exactly like that, yours will work too" is seriously not how you make robust and reliable software. It's a careless way of working where you basically don't want to attend to anyone else's needs. That's just not cool at all.

Imagine you being a developer since 80's, you learned C89 once and now you are in your sixties plus. Or you are a young developer which was raised by that old farts and you posess just the same knowledge. You are making tons of money working for Red Hat, Microsoft or whatever company today pays for contributing to Linux.

And now there are some kids coming, trying to introduce some language. Worse - they already wrote a big serious working driver in that weird language. Worst - that language immediately shows your magnificent code is buggy and needs changes.

And you start to feel loosing control because you don't understand shit. You start to smell retirement. You don't want to loose your warm place making money with 40 years old knowledge and protection instinct kicks in. Survival instinct even.

If that were 90s or 2000s you would lash on them insulting them, their kids, parents and pets, and overall saying they are bad developers and they should go to C kindergarten and summer camp. But today for that you can be easily cancelled so hard it would be much worse than simple retirement. So you start to create weird and stupid statements about the only thing you can insult - the language.

[mdedetrich]

It was just merged way too early. It shows the entitlement and "our way or the highway" mentality in the rust fanbase.
Writes a driver their way, gets told thats not how we do that here, goes crying on mastodon. Either you do it the established way or you properly explain why you want to do be the one special driver.

I believe it is best to remove the rust support for now, until at least there is a consortium based specification for the language and a good compiler implementation. It is just fully unacceptable that I now need GCC and LLVM just so I have one compiler being good at C and one compiler because I have no other choice.

I believe you have no fuken what you are talking about and you didn't even read the god damn article content.

Current existing C driver code is broken, Rust just revealed that. This is an issue completely irrespective of Rust, it's just an out of date maintainer with an unchecked ego issue that's the cause.

[Alexmitter]

I believe you have no fuken what you are talking about and you didn't even read the god damn article content.

Current existing C driver code is broken, Rust just revealed that. This is an issue completely irrespective of Rust, it's just an out of date maintainer with an unchecked ego issue that's the cause.

Thank you for your reply, this is exactly what I meant with my 11th word.