[spacelizard]

The echoes of Golang hype right now are somewhat ominous. A few years ago it was "concurrency primitives are going to change your life" and now it's "lifetime/ownership primitives are going to change your life."

I am somewhat disappointed that we still feel the need to create entirely new toolchains and rewrite everything just to support what should be an incremental improvement. But at a certain point, a lot of the reason for the hype is that it sells these solutions to newer developers that haven't had exposure to these concepts before. I suppose we can consider them fortunate for not having to do systems programming while suffering through implementing their own reference counting and garbage collection semantics.

[stcredzero]

A few years ago it was "concurrency primitives are going to change your life

Those, plus the total package of tradeoffs in Golang did for me, in terms of implementing a game server. The whole package really is game changing for what I'm doing.

I am somewhat disappointed that we still feel the need to create entirely new toolchains and rewrite everything just to support what should be an incremental improvement.

There's something wrong with programming as an entire field. Don't other fields figure out ways of not reinventing the wheel?

But at a certain point, a lot of the reason for the hype is that it sells these solutions to newer developers that haven't had exposure to these concepts before.

Luminaries of our field have been lamenting for decades that programming forgets its own history and discoveries. Maybe we should make it de rigeur that people relate their "inventions" with past art? A lot of times, when I point out past art, I get met with instant open hostility from younger devs. Is it any wonder that programming has the attributes of a popular medium, not a field of engineering study?

If young coders want to be the future intelligentsia and harbingers of a better kind of programming, they need to foster a set of subcultural norms that best leverages collective knowledge. Programming has to become a field that remembers its own history and can learn from its mistakes.

[bjz_]

Maybe we should make it de rigeur that people relate their "inventions" with past art? A lot of times, when I point out past art, I get met with instant open hostility from younger devs. Is it any wonder that programming has the attributes of a popular medium, not a field of engineering study?

[...]

If young coders want to be the future intelligentsia and harbingers of a better kind of programming, they need to foster a set of subcultural norms that best leverages collective knowledge. Programming has to become a field that remembers its own history and can learn from its mistakes.

Graydon has consistently pointed out Rusts influences, indeed it is one of the reasons behind its name - ie. 'rusty old ideas'. But the influences of the creator are not enough. We need to promote the value of a historical perspective in our discourse, and so I I'm glad for your post. As a young developer I see it in my peers, but even sadder I also see it in my mentors and elders. So often there is two extremes of fervent excitement and lack of imagination that things could be better beyond ones own experience... both borne out of a lack of historical literacy.

That said, I don't mind reinventing the wheel in new ways - every time our wheels get better and better. But perhaps they could get better faster with fewer steps backwards with a little schooling on the foundations of our field.

[stcredzero]

That said, I don't mind reinventing the wheel in new ways - every time our wheels get better and better.

Not always. Pry debugging Ruby can be pretty awesome but still often leaves a lot to be desired compared to the VisualWorks Smalltalk debugger+environment. Then there's the One Laptop Per Child project. How many developers were trying to recreate things that already existed in Smalltalk? What if they could've freed up the developer power wasted in that wheel reinvention? Seriously, a lot of the advanced state saving/rollback features would've just come for free in Smalltalk.

[coldtea]

>Not always. Pry debugging Ruby can be pretty awesome but still often leaves a lot to be desired compared to the VisualWorks Smalltalk debugger+environment. Then there's the One Laptop Per Child project. How many developers were trying to recreate things that already existed in Smalltalk?

How many Smalltalks were not closed source and with commercial IDEs and toolsets?

[stcredzero]

Squeak. That would've been enough. Smalltalk doesn't have an IDE in the traditional sense. The browser is a very small, thin app on top of the meta level and libraries.

[coldtea]

Squeak only got viable a decade or so after the demise of commercial Smalltalks.

[pjmlp]

Back in those days we just bought our tools.

[ZenoArrow]

>"That said, I don't mind reinventing the wheel in new ways - every time our wheels get better and better. But perhaps they could get better faster with fewer steps backwards with a little schooling on the foundations of our field."

You may be interested in Graal/Truffle. One of the main aims is to make it a playground for experimenting with programming language design by reducing the work that goes into the underpinnings of these new languages. It's also meant to offer fairly decent performance.

https://medium.com/@octskyward/graal-truffle-134d8f28fb69#.n...

https://github.com/graalvm/truffle/blob/master/README.md

[rch]

> There's something wrong with programming as an entire field

Look at all the research and prototyping in aviation. What's weird about programming is the amazing longevity of experimental ideas vs. heavily engineered products.

[stcredzero]

What's weird about programming is the amazing longevity of experimental ideas vs. heavily engineered products.

I like that very much! Also strange are the obscurity and outright myths surrounding heavily engineered products, followed by hyped reinvention.

[coldtea]

>There's something wrong with programming as an entire field. Don't other fields figure out ways of not reinventing the wheel?

No. In fact even in the actual field of "wheel manufacturing", we have been re-inventing the wheel since for ever.

Those who don't like "re-inventing the wheel", would they rather use stone or wooden wheels for their cars? Or perhaps some of the crude early metal ones? Or they rather use the re-invented modern mix of alloys and plastic?

[pjmlp]

I would say it is a matter of quality in computing related degrees.

My university did a pretty good job teaching the ways of the past.

Sadly things seem to have changed for the worst in the last 20 years.

[coldtea]

>My university did a pretty good job teaching the ways of the past.

That's because the past then was still modern.

[pjmlp]

Meaning teaching in the 90's the computing history of programming all the back to the early days.

Those majoring in systems programming had a good glimpse of OSes and programming languages.

[vvanders]

I think you'll find a lot of Rust's supporters actually come from a Systems Programmers background who've felt the pain before.

I've been doing C/C++ for close to 15 years now, I find that the borrow-checker not only helps with safety but also with guiding a good architectural foundation. Nothing keeps you from dropping into an unsafe block and cranking out very c-like code if you want as well.

All my green-field code now is Rust. Sure there's some hype but that's because people are genuinely excited about the language.

[spacelizard]

I have too but I don't see any magic in Rust's borrowing beyond what you would get with using shared_ptr, weak_ptr and move semantics in C++11. Yeah there is a certain class of errors that the Rust compiler will check that will still give you a segfault in C++. But it doesn't get around the programmer still needing to have a firm understanding of RAII and the difference between the stack and heap in order to be effective in either Rust or C++, or Golang, or C99, or what have you.

[vvanders]

Rust is actually much closer to unique_ptr than shared_ptr or weak_ptr. The use of Rc is pretty small in practice.

The thing is that even with unique_ptr I can easily reach in via get() and I've broken the shared mutable state contract. It's trivial for me to trace what modifies things in Rust, I just follow the 'mut' annotations and since only one thing can hold it at a time it's very straightforward.

Rust also has one the best package managers I've seen. Union types which are the best way to represent states and state machines bar none. An iterator class that's the best I've seen(seriously has anyone actually used <algorithm>? How man begin/end pairs do I need?).

Plus the community is super-welcoming, there's more reasons but those are my top ones right now.

[coldtea]

>But it doesn't get around the programmer still needing to have a firm understanding of RAII and the difference between the stack and heap in order to be effective in either Rust or C++, or Golang, or C99, or what have you

Who said it does? And why would it have to?

It'a just about catching those "certain class of errors that the Rust compiler will check that will still give you a segfault in C++".

Plus a nicer looking language, with a more modern design.

[dbaupp]

The more strict static checking in Rust means that the compiler can actually be a teaching tool for concepts like RAII and the heap/stack difference: it flags and explains places where the programmer is doing that doesn't make sense in those settings. A lot of effort has been put into the compiler diagnostics to make this compiler-is-a-teacher concept a reality.

[pjmlp]

The magic is that there is no way to enforce that in modern C++, in spite of current attempts.

There will always be that compiler or team that doesn't have any kind of reviews, static analysis, making use of binary libraries or legacy code.

Hence why safety at language level is much better than third party tooling, as tools can be ignored.

[steveklabnik]

For one, additional safety. Rust's equivalent are more safe. Secondly, speed. Rust has several advantages: with shared_ptr, for example, you have the Rc/Arc split, the decreased number of refcount bumps, etc. (some of that speed comes from the safety, even...)

[steveklabnik]

Rust is aiming for more than an incremental improvement. Take Servo, for example. The goal is to provide a generational leap in rendering engines, and while it's not done yet, the initial results are promising.

I agree that "salvation" is a bit much, though.

For whatever hype Go had and may have, it's still a resoundingly successful language that's building meaningful stuff that improves people's lives. Sounds like a success to me.

[to3m]

Move semantics, man.

I don't personally divide my life into "before move semantics" and "after move semantics" - but thanks to move semantics, I have to anyway. And I feel like I'm all the better for it.

[coldtea]

>I am somewhat disappointed that we still feel the need to create entirely new toolchains and rewrite everything just to support what should be an incremental improvement

Any ideas about how to add this small "incremental improvement" of Rust-like safety to C, for example?

[jstewartmobile]

If we were betting money, I'd put mine on Golang over Rust just because of the piles of money and effort Google keeps pouring into it. Rust is a little cooler, but I just don't see Mozilla having enough resources to keep-up the momentum long-term.

[binarycrusader]

I don't see how Google is really putting piles of money into Go. In fact, in terms of maintainership and porting I'd say collectively that many other companies are investing more in Go than Google. What can you point to as piles of money?

[jstewartmobile]

Perhaps not that much by valley standards, but I'm sure just the salaries of Rob Pike and Brad Fitzpatrick alone would amount to more love than a lot of languages get.

[mmstick]

Go's not been receiving any piles of money from Google. Compare Go's API documentation to Rust's API documentation. Where's the `cargo` and `rustup` equivalent for Go? Compare the number of Go packages to the number of Rust packages. Seems Go's not doing so great despite being stable for much longer.

> I just don't see Mozilla having enough resources to keep-up the momentum long-term.

Did you forget about Samsung? Mozilla and Samsung aren't the only two companies backing Rust.

[jstewartmobile]

By TIOBE's stats, Go's popularity is up 1.74%. Rust's is up 0.316%.

http://www.tiobe.com/tiobe-index/

Call it money or whatever, but Go seems to be getting substantial TLC from the people working on it for every point release.

Of course, this could just be an easier thing for them to do since their approach is not as ambitious as Rust's.

[dbaupp]

When was the last time you heard of Samsung being involved with Rust as a company? I'm fairly sure that is no longer a thing.

[nine_k]

Unfortunately, golang is just a different league.

[kmicklas]

These piles of money don't seem to be amounting to much because rust is still an objectively far superior language.

Google is not a language company and doesn't hire good language people so I don't see this changing.

Disclaimer: I work for Google

[jstewartmobile]

No argument on superior, it's just that superior doesn't win the day. Look at C++. Look at JavaScript. It would be surprising if the industry made the better choice for a change.

[eropple]

Hmm--with some edge-case exceptions that never got a strong push outside of some niches (Ada comes to mind, where it achieved some traction in DoD circles), I wonder if C++ wasn't the superior choice given its time and place. By no means is it perfect, but in the early to mid 90's there weren't many competitors with a similar feature set that had anybody pushing hard behind it. To my mind it seems not dissimilar.

If I have to write native code, right now it'll be C++ precisely because it's able to express complex problems with a high degree of getting-it-right (because while I would never claim to understand C++, I feel confident in my understanding of the subset of C++ that I use). I really like Rust and look forward to it being a better fit for what I want to do, and every time I look at it it's getting closer.

[kmicklas]

Spot on, C++ has always been the best systems language for real world projects until Rust came along.

[kmicklas]

Yeah but unlike, say, Haskell, Rust seems to have good enough momentum from the right crowds.

[foota]

But rust reuses some tooling, for instance, llvm.