Comment by mustache_kimono | Hacker News Reader

Comment by mustache_kimono | Hacker News Reader

Parent Full thread

mustache_kimono

> The article explicitly states

Again -- I think your impression is the result of the contentious tone of the article. Yes, the article explicitly states:

    "Think of programming languages as belonging on a continuum, with assembly at one end and the interface to the Starship Enterprise’s computer at the other. Low-level languages are “close to the metal,” whereas high-level languages are closer to how humans think."

But then spends the rest of the article debunking this commonly held notion, specifically and explicitly re: C, but also implicitly re: assembly.

See the very next section "FAST PDP-11 EMULATORS"

    "The root cause of the Spectre and Meltdown vulnerabilities was that processor architects were trying to build not just fast processors, but fast processors that expose the same abstract machine as a PDP-11. This is essential because it allows C programmers to continue in the belief that their language is close to the underlying hardware."

The author obviously knows that assembly suffers from the same abstraction penalty. The author is saying, because C and processor design has been so tightly intertwined, we cannot program "close to the metal" because "the machine" is actually a very fast PDP-11 emulator.

See also the section "IMAGINING A NON-C PROCESSOR", where the author explicitly discusses alternative processor designs (which would of course require new assembly languages!).

The author is actually trying something like a reductio on your mental model. When the author states "Think of programming languages as belonging on a continuum", the author is really saying "This is everyone's impression, but ... when you look a little deeper you see the cracks (which are actually contradictions)."

At my most charitable, I would give the author credit for knowing that assembly is not low-level by his own definition, but nonetheless intentionally misleads readers who do not know that assembly does not expose speculative execution instructions any more than C does. If the author knew this, and his central premise were that no low-level programming language existed anymore, the article would be titled differently and he wouldn't be making the arguments against C specifically. Really, that would be an entirely different article if written honestly. But this is essentially packaged as clickbait that takes advantage of ignorant readers to make the author sound smarter and more insightful than he really is, and still contains logical errors perpetuated as misinformation to the readers who don't know better. After writing out my charitable interpretation of the author's capabilities, this interpretation only leaves me more disgusted with the article as a writing output.

> nonetheless intentionally misleads readers

I'm really not certain that's the idea, and it certainly does not feel very charitable. Perhaps you are holding on a little too tightly to this high vs. low level distinction (the simple mental model being attacked)? I compared the author's argument to a reductio. A reductio is not intentionally misleading?

This paper reminds me of "It's Time for Operating Systems to Rediscover Hardware". See: https://www.youtube.com/watch?v=36myc8wQhLo

There, an argument is made that our simple model of "the machine" is also wrong. There, the speaker points out much of the software that is running on our complex SoCs, with multiple cores, is firmware. To which, I'd imagine you might argue: "But that firmware is not the OS?! This talk is misleading!"

> If the author knew this, and his central premise were that no low-level programming language existed anymore, the article would be titled differently and he wouldn't be making the arguments against C specifically.

I am not sure. I believe the reason C is targeted specifically is because C communities are where this myth, and its religiosity (!), is the strongest.

> But this is essentially packaged as clickbait

I'd agree that the article is provocative, but it would seem to have good reason to be. Lots and lots of people think both C and our processors must work one way. That there is or was some level of naturalism/determinism at play. The author is simply pointing out -- not so much.

> After writing out my charitable interpretation of the author's capabilities, this interpretation only leaves me more disgusted with the article as a writing output.

Yes, it was designed to make you mad. But if you were forced to write a rebuttal to the entire article, from a charitable POV, I think you'd see there is some value to the reader in realizing this tight coupling (C and processor design) is not a necessary condition.

The idea that processor evolution is blocked by the need to run C code is wrong. We had many alternative designs, both for languages and also for processors. They were simply not successful. I still remember the pain of segmented memory in 286. Also much of the criticism in the article would apply to von Neumann / Harvard architectures in general, and is not specific to PDP-11 and C.

Then also, where new programming methodologies such as CUDA are invented to allow new processor designs, it turns out that they can be quite successful despite moving away from C. But then it also turns out that this programming model was actually not that great to program in, and people try hard to move back again by making the hardware more capable.

> I believe the reason C is targeted specifically is because C communities are where this myth, and its religiosity (!), is the strongest.

C being a low-level language is not a myth if, as normal people do, you consider assembly languages to be low-level languages. The article itself offers assembly as the low-level language and makes no effort to redress this later.

If you want to argue that there are no low-level languages when it comes to programming modern CPUs, you are free to do so, but that is a different argument. And it is arguably not a fruitful one because then the terminology loses all meaning. Okay, you've defined assembly as a high-level language. Now what have you accomplished other than making it harder for people to articulate and categorise classes of languages? We'll still need an adjective for distinguishing between such wildly distinct languages assembly and Python, so we have to come up with something else to replace "low-level" and you haven't accomplished much of anything at all.

If you think low-level programming has gotten too far from the bare metal, or if you think processors are designed for C and that's a problem, just argue that directly instead of this really torturous detour singling out C and how people use terminology to communicate useful concepts.

> C being a low-level language is not a myth if, as normal people do, you consider assembly languages to be low-level languages.

I'd suggest you're holding that stick too tight!

> If you want to argue that there are no low-level languages when it comes to programming modern CPUs, you are free to do so, but that is a different argument.

Not if you read the whole article?

> And it is arguably not a fruitful one because then the terminology loses all meaning. Okay, you've defined assembly as a high-level language.

Again, what if the terminology isn't very important? I'd argue the distinction between high and low level languages is not an important distinction, because it is so crude. For example, C/C++/Rust have all been described as high level languages at one time or another. C people seem to be the only ones that take real offense to this.

Reply on news.ycombinator.com