Rustls Multi-Threaded Performance Is Battering OpenSSL

[madinside]

Did you hear about the C++ std committee UB question case?

Isn't this here about Rust and not about C++?

[tesfabpel]

The thing is the other implementations have decades of baggage. They probably can't touch or understand a lot of the code.
We will have to see how well rustls stands up after 30 years of different people hacking on its code base.

Thing is, Rust is easier to confidently refactor thanks to the compiler's forced checks and a language that allows those checks. It allows you to work without the consistent cognitive load of trying to enforce C / C++'s invariants to not cause UB / crashes.

[fotomar]

They are just competent coders freed from dealing with harassment. What's not to like? Isn't it about merits?

if they're semi-anonymous people behind a screen, what would cause them to get harassed?

[Britoid]

OpenSSL is written in C, Assembly, and Perl. Those are all known for their shittiness in multi-core contexts. Perl is slow as hell. It's sad that the militant pink-lightblue-white coding sock community is winning because they're openly banning people from asking questions like in the std c++ committee. But it's great to see these piece of shit languages are slowly but surely dying. They're not really designed for modern post-2005 computers, anyways.

Winning? Trump won

[Anux]

Did you hear about the C++ std committee UB question case?

Holy shit, what's wrong with people these days? Can this get any more ridiculous? I mean if there is one woke psycho in the committee the others should be intelligent hackers and oppose this bullshit right away.

I guess someone can't wait till C++ becomes naturally irrelevant.

The thing is the other implementations have decades of baggage. They probably can't touch or understand a lot of the code.

Of course if you mix C/C++/ASM and even Pearl it's likely to end up that way. But if you look at BoringSSL it has perfect scaling just the overall performance lacks behind. It all comes down to code practice and documentation.

We will have to see how well rustls stands up after 30 years of different people hacking on its code base.

Good thing with rust is that it (somewhat) forces you to code well and you can have build in unit test to spot any regressions.

[TheMightyBuzzard]

Good on them. Tell them to go write their own shat in Rust instead of screwing with existing projects and they did. And it looks pretty good. Of course they're rolling their own crypto and we should all know how that saying goes.

So, can you use it? Sure. Is it mature and tested enough to use it for your illegal world domination communications? Well, you can if you like but wise people won't be doing so.

Holy shit, what's wrong with people these days? Can this get any more ridiculous? I mean if there is one woke psycho in the committee the others should be intelligent hackers and oppose this bullshit right away.

I guess someone can't wait till C++ becomes naturally irrelevant.


Of course if you mix C/C++/ASM and even Pearl it's likely to end up that way. But if you look at BoringSSL it has perfect scaling just the overall performance lacks behind. It all comes down to code practice and documentation.

Good thing with rust is that it (somewhat) forces you to code well and you can have build in unit test to spot any regressions.

No, it forces you to not run afoul of the most obvious footguns, it does not make you code well. Not even somewhat.

[cynic]

It's sad that the militant pink-lightblue-white coding sock community is winning because they're openly banning people from asking questions like in the std c++ committee

well, the Rust community is not much better in that regard: it's full of people with couloured hairs and pronouns in bios.
also, did you forget about the recent Rust Foundation dramas?

[oleid]

This is where Rust shines. Boring TLS is C++/Assembly/C, OpenSSL is C. Rustls is 100% Rust,.

That's not entirely true. You can use 100% rust in rustls, however, there are also configuration variants where certain crypto primitives use well-tested assembly implementations.
As stated in the benchmark, "This was used with aws-lc-rs". So Amazon's crypto primitives were being used.

Still, the multi-core performance is quite nice.

[gufide]

Did you hear about the C++ std committee UB question case?

You have no idea what happened there so stop writing stupid

First, the committee cannot ban people. They have actual sex offenders operating there and no willingness or the means of banning them, even though many incredibly talented engineers had to quit because they felt deeply uncomfortable being forced to interact with him. If you're in the committee people are forced to interact with you and address your criticism, for the better or worst.

Then for that particular case, the entity sponsoring that person banned him from being sponsored by them again not just because of a complaint, but his refusal to address the complaint. I do agree that the nature of the complaint is questionable at best, but it was a stupid hill to die on. He could have turned the controversy against them instead.

[cynic]

You have no idea what happened there so stop writing stupid

First, the committee cannot ban people. They have actual sex offenders operating there and no willingness or the means of banning them, even though many incredibly talented engineers had to quit because they felt deeply uncomfortable being forced to interact with him. If you're in the committee people are forced to interact with you and address your criticism, for the better or worst.

Then for that particular case, the entity sponsoring that person banned him from being sponsored by them again not just because of a complaint, but his refusal to address the complaint. I do agree that the nature of the complaint is questionable at best, but it was a stupid hill to die on. He could have turned the controversy against them instead.

we have a exact idea of what happeneded, and Andrew was perfectly right.
to those who missed it: http://tomazos.com/ub_question_incident.pdf