Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

A Sad Day for Rust

words.steveklabnik.com

Post reply on HN
Print view View original

991–1000 of 1001 posts

Re: A Sad Day for Rust

#991

Post by namibj »

_-___________-_ wrote: post #970
namibj wrote: post #892

Earlier quoted context omitted.

I'm curious what you'd recommend instead for async pgsql requirements as a Rust web framework. Actix seemed interesting, but I didn't know about the masses of unsafe. These too eliminate it, because it'd be able to crash/break a cluster that uses deliberate unsoundness for its serialization/deserialization needs in the vain of speed (timely-dataflow), allowing memory issues to spread (at least in theory)... I thus ca…

We use tokio-postgres. Previously we had our own connection pool implementation, but we recently switched to deadpool.

Thanks! The pipelining is just really important to get proper performance out of CockroachDB (mostly due to network and coordination overhead due to it's cluster architecture).

Re: A Sad Day for Rust

#992

Post by Cladode »

Animats wrote: post #988
Cladode wrote: post #978

Earlier quoted context omitted.

there are things that you either can't express, or are very hard to express, in safe Rust. This is true (and in some sense inevitable by Rice's theorem) but it is unclear what can be done about this. Bear in mind that Rust's design constraints are - decidable (indeed fast) type-inference/type-checking - simple typing system - you don't pay for features you don't use It is easy to get rid of unsafe blocks if you throw…

Oh, not Rice's theorem again. That only applies to infinite systems. There are useful decidable properties for a reasonably large set of programs. Yes, they're "uninteresting" mathematically, but we're talking about memory safety here. If your program is anywhere near undecidable in that area, you're doing it wrong. Any time you're close to the edge on that, you can get decidability with a run-time check. The Microso…

I'm still unclear about this, should this be done by the typing system or by an external static verifier:

   simple inductive proofs to show 
   that as you initialize 
   another element and add 1 to 
   the counter
If the latter, you can already do this with Rust as it is, it's just a question of building such a static verifier. If the former, I really doubt that this can fit in Rust's typing system without major surgery. We know well how to add arithmetic to typing systems, but the natural approach to doing so (dependent types) will probably make type-checking / -inference undecidable.

Re: A Sad Day for Rust

#993

Post by aws_ls »

thu2111 wrote: post #969
akamaka wrote: post #549

Earlier quoted context omitted.

I’m replying to your message using software written in C, so do you think that it might be a little bit unfair to call it “broken”? I suppose I’d agree if you’re making a general argument that most of the software we use is broken to some degree, but it has an entirely different meaning to directly tell someone “Your library is broken”.

Are you? All web browsers I know of are written in C++ and always have been. In modern C++ you don't use char* to mean "pointer to byte array of unknown size", you use std::vector. Moreover Chrome actually uses garbage collected C++ (they call it Oilpan).

They may be using Linux as OS.

Re: A Sad Day for Rust

#994

Post by fluffything »

rowanG077 wrote: post #971
fluffything wrote: post #949

Earlier quoted context omitted.

This is how it works in practice for the Rust compiler for the Rust standard library, and for a lot of foundational crates in crates.io. (Pretty much every well reviewed crate in cargo crate review either does this, or does not contain any unsafe code at all). We also have tools that change for this for very large projects (e.g. cargo-geiger), and tools that help you test your proofs (e.g. cargo-miri). For some unsaf…

I didn't say it no one will do it. I said almost no one. The Rust compiler is about as far from a normal project as can be. I would like to see this changed but really I have never even seen a proof requirement when looking at Rust jobs. Not even once.

> I would like to see this changed but really I have never even seen a proof requirement when looking at Rust jobs. Not even once.

?

Any B.Sc. in CS can do most of the unsafe proofs in a one liner. All crates I maintain require unsafe blocks to be commented with a proof, for most of them the proofs are trivial, and for all of them that weren't, the unsafe code was correct, and the correct one had a trivial proof.

Re: A Sad Day for Rust

#995

Post by amyjess »

infinity0 wrote: post #961
Traster wrote: post #960

Earlier quoted context omitted.

> initially were very nice and tried to give the author very polite suggestions on improvement, and it was only after the author being extremely dismissive[1], not admitting to his own flaws, not wanting to learn from others, and abusing his own power in shutting down valid discussion, that things turned nasty. Yes, nice comments such as >seriously? Please just stop writing Rust. You do not respect semver, you do not…

That was after the author was extremely dismissive with "this patch is boring". Seriously, who says that??

And that still doesn't justify harassment. If that person said any of these things to the maintainer on HN, the mods would shut them down and for good reason.

Re: A Sad Day for Rust

#996

Post by rowanG077 »

fluffything wrote: post #994
rowanG077 wrote: post #971

Earlier quoted context omitted.

I didn't say it no one will do it. I said almost no one. The Rust compiler is about as far from a normal project as can be. I would like to see this changed but really I have never even seen a proof requirement when looking at Rust jobs. Not even once.

> I would like to see this changed but really I have never even seen a proof requirement when looking at Rust jobs. Not even once. ? Any B.Sc. in CS can do most of the unsafe proofs in a one liner. All crates I maintain require unsafe blocks to be commented with a proof, for most of them the proofs are trivial, and for all of them that weren't, the unsafe code was correct, and the correct one had a trivial proof.

It's fine that YOU do it. But how many unsafe blocks have a proof in the entire Rust ecosystem (read all crates). 5% maybe? I'm not sure what your point is that any B.Sc. in CS can do it.

Re: A Sad Day for Rust

#997

Post by strbean »

pjkundert wrote: post #724
strbean wrote: post #656

Earlier quoted context omitted.

I have to question your position from a moral standpoint though. If you were a rollercoaster engineer, and you saw that a rollercoaster had an unsafe design, would you follow a similar approach? "I'm not going to ride that, but I'll let this line of people ride it without warning them." Obviously the stakes are wildly different, but still...

Look up “Ergodicity”...

Okay... huh?

Re: A Sad Day for Rust

#998

Post by comfyinnernet »

twic wrote: post #976
atoav wrote: post #955

Earlier quoted context omitted.

To me the maintainer sounds more like the boy who brought everybody cake and then left the party when the other boys started nagging about how their mothers would make much better cake and how his cake is all wrong.

It's more that he bought a cake, but it wasn't cooked in the middle, so there was a chance people could get food poisoning, and the other people at the party pointed that out and suggested ways to cook cakes more consistently, but the author really likes cooking soggy cakes, and he's never got sick from it so it must be fine, so he left the party and told everyone the others were mean to him.

It's more like he bought a cake but the brakes were faulty and it skidded off the road and killed an innocent trifle.

Re: A Sad Day for Rust

#999

Post by pjkundert »

strbean wrote: post #997
pjkundert wrote: post #724

Earlier quoted context omitted.

Look up “Ergodicity”...

Okay... huh?

Failures of Actix will be “ergotic”; individuals experiencing the failure will be able to play again, experiencing the long-term average failure rate.

Failures of a roller-coaster are “non-ergotic”; individuals never experience the long-term average failure rate, since they’re, you know, dead...

This has got to be the most common and tragic type of mis-application of statistics I’m aware of; I know I used to do it all the time! Virtually nobody teaches/talks about it, but it is arguably more critical to understand than “causation vs. correlation”, or “post hoc ergo propter hoc”.

Re: A Sad Day for Rust

#1000

Post by infinity0 »

amyjess wrote: post #995
infinity0 wrote: post #961

Earlier quoted context omitted.

That was after the author was extremely dismissive with "this patch is boring". Seriously, who says that??

And that still doesn't justify harassment. If that person said any of these things to the maintainer on HN, the mods would shut them down and for good reason.

Saying someone's "patch is boring" is also harassment.
Post reply on HN