5 ms·

The description of Rust in this blog post is not accurate in my opinion. 1. Compile times aren't great right now, but 20s for 2400loc is a real anomaly, and im

The description of Rust in this blog post is not accurate in my opinion. 1. Compile times aren't great right now, but 20s for 2400loc is a real anomaly, and improving compile time is a focus of the team now that the language is stable. This is the only specific painpoint the blog post mentions with Rust, and its just a single anecdote. 2. The idea that Rust is a premature optimization is a weird meme. Faster, safer, more expressive. Rust is at least 2/3 for pretty much every 'mainstream' language. I would rather write Rust than any other language I know, and I would be more productive in it. 3. Rust is not a Flub language struggling to overcome a niche - not yet. Rust is a language that became stable 3 weeks ago and time has not yet told whether it will be 'mainstream' or not. Rust has a memory model that includes some degree of manual management, like C (instead of malloc/free and pointer arithmetic, you have ownership and references). This makes it a poor first language, like C. But no one says C is a language for the 'elite' because you have to think about the memory model.

stevedonovan 11y ago
It was perhaps unfair to single Rust out given that it is so young, but there are costs to any innovation, and the first cost that leaped out at me was stories of long compile times. Besides, extraordinary claims require extraordinary evidence. Personal anecdotes of increased productivity remain anecdotal - it may be more a property of the programmer than the language. I don't personally have a horse in the race, but pointing out cost points is needed.
- jerf 11y ago
  I really don't expect Rust to exhibit what most people mean by "increased productivity", ever. Most people mean something fairly local by that, or at most medium sized, that it allows them to take some task and write code faster than other languages that accomplishes the task. Rust is always going to labor under the disadvantage that it holds you to a higher degree of correctness than other languages, and at best it can tie with another language that doesn't hold you to those guarantees of correctness. Where it'll increase productivity or not is at larger scales. If Servo comes out at production quality in some reasonable time (call it 2 years or less from now, given where they are), able to function on the general web most of the time, and it's also faster than current renderers by a significant margin, and with some small fraction of the development effort (probably even if you include the entire Rust development effort; do not underestimate the effort poured into the existing engines!), I will consider Rust's story proved. We've now had many people take cracks at writing renderers, in the fastest language of the time for the task (C++), and I daresay they've converged at fairly similar levels of performance with little reason to believe the renderers are going to get much faster. Said renderers have had massive optimizations applied to them, and this Servo that I'm expecting to be much faster will actually be version 1.0, basically. If Rust is to work at all, there will very, very rapidly be a class of development that it will be actually insane to write in anything but Rust... but that will only be the very high-and-large end. (It will take many more years for this fact to penetrate through to the development community at large.)
  - Dewie3 11y ago
    The standards one holds emerging technologies to tends to be proportional to how much one personally dislikes them. (and of course the corollary vice versa)
WalterGR 11y ago
The idea that Rust is a premature optimization is a weird meme. I don't think the sentence you're referring to is really about Rust per se, but rather the author's feelings about manual memory management. Rust just happened to be the language being discussed at the time: "To use Rust for writing programs that can afford the overhead of garbage collection would be premature optimization." The idea being that manual memory management is such a productivity black hole that the lack of GC in a language dominates any 'Flubby' features the language provides.

stevedonovan 11y ago

It was perhaps unfair to single Rust out given that it is so young, but there are costs to any innovation, and the first cost that leaped out at me was stories of long compile times. Besides, extraordinary claims require extraordinary evidence. Personal anecdotes of increased productivity remain anecdotal - it may be more a property of the programmer than the language. I don't personally have a horse in the race, but pointing out cost points is needed.

jerf 11y ago
I really don't expect Rust to exhibit what most people mean by "increased productivity", ever. Most people mean something fairly local by that, or at most medium sized, that it allows them to take some task and write code faster than other languages that accomplishes the task. Rust is always going to labor under the disadvantage that it holds you to a higher degree of correctness than other languages, and at best it can tie with another language that doesn't hold you to those guarantees of correctness. Where it'll increase productivity or not is at larger scales. If Servo comes out at production quality in some reasonable time (call it 2 years or less from now, given where they are), able to function on the general web most of the time, and it's also faster than current renderers by a significant margin, and with some small fraction of the development effort (probably even if you include the entire Rust development effort; do not underestimate the effort poured into the existing engines!), I will consider Rust's story proved. We've now had many people take cracks at writing renderers, in the fastest language of the time for the task (C++), and I daresay they've converged at fairly similar levels of performance with little reason to believe the renderers are going to get much faster. Said renderers have had massive optimizations applied to them, and this Servo that I'm expecting to be much faster will actually be version 1.0, basically. If Rust is to work at all, there will very, very rapidly be a class of development that it will be actually insane to write in anything but Rust... but that will only be the very high-and-large end. (It will take many more years for this fact to penetrate through to the development community at large.)
- Dewie3 11y ago
  The standards one holds emerging technologies to tends to be proportional to how much one personally dislikes them. (and of course the corollary vice versa)

WalterGR 11y ago

The idea that Rust is a premature optimization is a weird meme. I don't think the sentence you're referring to is really about Rust per se, but rather the author's feelings about manual memory management. Rust just happened to be the language being discussed at the time: "To use Rust for writing programs that can afford the overhead of garbage collection would be premature optimization." The idea being that manual memory management is such a productivity black hole that the lack of GC in a language dominates any 'Flubby' features the language provides.