Hacker News Reader

Comment by pdimitar | Hacker News Reader

So, we have to choose between...

- A tiny community of hobbyists willing to support niche architectures that have zero relevance to any mainstream computing,

- Embrace a newer, stricter ecosystem with more guarantees and clearer communicated support tiers that's also constantly improved upon by a big number of both dedicated volunteers who donate their time and effort, and paid professionals. The only tradeoff: it supports less architectures. For now.

Am I understanding the article correctly?

If so, I am definitely in favor of the latter, and I think many others are as well.

View on HN

It's interesting that the community praising multiculturalism favors hegemony of only a few computer architecture. Portability used to be paramount of design...

Most CPU architectures are "fake diversity"; for example both Alpha and ARM64 are 64-bit little-endian with a weak memory model. Sure, S/390 is 31-bit and supports BCD while in PA-RISC the stack grows upwards and IA-64 is VLIW, but these are trivia that are not comparable to the diversity of human cultures. For decades programmers have wasted their time porting software to different-but-not-better architectures, mostly for the benefit of the vendors who fragmented the market in the first place instead of standardizing.

> For decades programmers have wasted their time porting software to different-but-not-better architectures

Microarchitecture, register layout, ABI also constitute differences which have real-world uses, Not to mention sheer competition to avoid architecture-rot. Only targeting intel & ARM opens yourself to problems, cf. the upcoming nVidia hell ARM is about to experience.

Just because Rust-preaching (without practicing it, of course) Starbucks-spipping average HN readers don't know about it doesn't make it inexistent or necessarily wrong-think.

> Microarchitecture, register layout, ABI also constitute differences which have real-world uses

They do. I started coding on a 6502-based machine and used machine code to find prime numbers, some 27 years ago. I've used 1-2 other non-mainstream CPUs (whose names I don't even know) before diving neck-deep into the mainstream. It was fun, absolutely. It has potential, absolutely. Was it realized? Nope.

However, I can't resist but asking: if those things do have their uses then why didn't the hobbyists support them through patches to GCC / clang and LLVM?

Don't get me wrong. If you tell me we are stuck in a local maxima in CPU architectures, I'll immediately agree with you! But what would you have the entire industry do, exactly? Business pays our salaries and they need results in reasonable timeframes. Can you tell the guy who is paying you: "I need 5 years to integrate this old CPU arch with LLVM so we can have this feature you wanted last month", with a straight face?

> Just because Rust-preaching (without practicing it, of course) Starbucks-spipping average HN readers don't know about it doesn't make it inexistent or necessarily wrong-think.

That is just being obnoxious and not arguing in good faith. Example: I do use Rust, although not 100% of my work time.

You should try the Rust language and tooling -- and I mean work actively with it for a year -- and then you could have an informed opinion. It would make for a more interesting discussion.

Do I like how verbose can Rust be? No, it's irritating.

Do I like how cryptic it can look? No, and it wastes time mentally parsing it (but it does get better with time so 50/50 here).

Does it get stuff done mega-quickly and safer than C (and most C++)? Yes.

Does it have amazing tooling? Yes.

Does it get developed more and more and serve many needs? Yes.

Does it reduce security incidents? I'd argue yes although I have no direct experience. Memory safety is definitely one of the largest elephants in the room when security is involved.

---

You have a very wrong idea about the average Rust user IMO. I don't like parts of the whole thing but it helped me a lot several times already -- and it gave me a peace of mind. And I've witnessed people migrating legacy systems to it and showing graphs in meetings that demonstrate that alarms and error analytics percentages plunged to 0.2% - 2% (and they were always 7% - 15% before).

Just resisting something because it starts going mainstream is a teenager rebellion level of attitude and it's not productive. Do use Rust yourself a bit. Then you can say "I dislike Rust because of $REASON" and then we can have a much more interesting discussion.

> However, I can't resist but asking: if those things do have their uses then why didn't the hobbyists support them through patches to GCC / clang and LLVM?

They didn't decide to create a whole new language and make everything dependent on it.

At some point, when you reach a critical mass, you have to spend more on seemlessly "irrelevant" tasks, like supporting other architecture. Don't shift the problem away by making ridiculing it, own your shortcomings.

Okay, that's a more fair and balanced point of view.

However, let's not forget one of the main points of original article: nobody promised those people that their dependency's dependencies will never change. The crypto authors made a decision to go with Rust. If dependents want to continue using it, they have to adapt or stop using it.

As I've said above: backwards compatibility is an admirable goal but it doesn't override everything.

> As I've said above: backwards compatibility is an admirable goal but it doesn't override everything.

You'll never get a job at either Microsoft, or in any system jobs where backward compatibility is paramount (say, the Linux kernel) for millions, if not billions. Just going the Apple "fuck you" way is arrogant at best, disillusioned at worst, especially when you're an irrelevant language.

I don't think the discussion will ever get anywhere if we only compare polar opposites.

I'm not advocating for either extremity, what about you?

Backward compatibility is, by definition, an all or nothing binary deal. You can't have it otherwise.

[and yes, this is gonna be unpopular in a post-modernist era where everything get constantly redefined and where there is no such thing as "meaning".]

Sorry that your work has made you so frustrated. It sounds stressful. IMO you should consider exiting your current company or area. Judging by your comments, you are pretty jaded (and set in your ways).

I am not interested in discussing extremes as mentioned in two separate sub-threads now but you do sound like you need a break. Good luck, man.

Everybody can do whatever they like with their time. But if they want their exotic CPU architectures then they can support them themselves, no?

However, demanding mainstream tools to lag behind because of said exotic architectures is unrealistic. At one point we all want to progress and advance our craft, especially the one that pays our bills.

I'm not against multi-culturalism. But we can't have it at the expense of everybody else outside your small bubble having their tooling hampered and/or lagging behind on features that a lot of us need for commercial work (and not only, I'd argue).

Backwards compatibility is like everything else: it can't be praised as an absolute value and damn everything else.

> However, demanding mainstream tools to lag behind because of said exotic architectures is unrealistic

Rust is mainstream on HN, not in the real world. Because "Uber" or whatever unicorns uses it doesn't make you mainstream.

https://madnight.github.io/githut/#/pull_requests/2020/4 - Rust is barely reaching 1%.

I strongly disagree on those metrics and I would question their coverage with the real world out there.

Rust is getting more and more prevalent and I'm saying that as a person that has barely worked in only one SV company for the last 5 years.

I'm working outside the mainstream companies and I'm still seeing Rust gathering mindshare all the time wherever I go. And I'm not even hired for Rust positions.

Anecdotal for sure, I'll agree, but your observation is no less anecdotal than mine.

Rust brings very real advantages to the table and seeing people rebelling against it only on principle (and not on merit) is getting increasingly baffling. Feels like an emotional rebellion versus resistance based on facts and merit.

Show me a metric which makes Rust a relevant language.

> Anecdotal for sure, I'll agree, but your observation is no less anecdotal than mine.

I haven't seen Rust in any "top X language" news. Prove me wrong.

Do you participate in any language popularity study that you stumble upon? I know I don't. Add 99% of all my colleagues ever don't as well.

Even if I found a study that corroborated my observation I'd still not trust it. I don't make a habit out of supporting dubious studies only because they support my point of view.

From what I've seen for 19 years of career, most working programmers refuse to participate in such studies.

Hence I don't trust them either way. They work with a non-representative sample of the population. Not a big enough sample for the study to be valid.

Instead of shooting the messengers, please provide me a non-anecdotal reference about Rust being relevant.

This will not go anywhere. :)

I prefer to look around -- this has always been giving me much more objective info throughout my entire career.

I get your skepticism but you are not arguing in good faith. I already asserted that to me those language popularity contests are dubious and non-representative.

If you disagree with that premise then we have zero common ground and can't discuss the topic. ¯\_(ツ)_/¯

> This will not go anywhere. :)

Of course, you don't have any argument, so you're posturing and running away. Very immature.

Have you seen the stack overflow survey over the last few years? It has some interesting data.

https://insights.stackoverflow.com/survey/2020#technology-mo...

Reply on news.ycombinator.com