Hacker News Reader

Comment by phyzome | Hacker News Reader

It's hard for me to imagine using any new network-touching software written in C. We have safer options now...

Indeed. The written in C part of the headline is a red flag, rather than the advertisement it is probably intended to be. It basically translates to "a piece of software that must be thoroughly audited by highly specialized security experts every few months in order to be even remotely confident that it doesn't contain catastrophic vulnerabilities that allow arbitrary code execution by unprivileged remote attackers".

Given your extreme hyperbole, you probably aren't the target audience then.

For me, it translates to "a piece of software universally compatible with a vast amount of compilers and platforms that is probably trivial to compile even in 10 years".

Those are orthogonal concerns, and you're both right. But that absolutely wasn't hyperbole.

just don't. and I am waiting for your contribution in free and open source software to write correct, safe, and performant in Ada. It'd be much more welcome if you write the formal business requirements in Spark.

Rust? why would want to write SAFE software in a language without formal spec? I question the morality af those people.

I too question why people make flame-baiting comments about topics they don't understand.

agree, always see that kind of replies in threads related to software written in C. what those people want, eh? always derailing thread like that. I guess it's RESF in action. makes me despise Rust, tbh.

I was making a remark about your comment.

And it's because most security vulnerabilities these days have to do with memory issues, which must be manually enforced in languages like C. In Rust you get most of that for free, often even more optimized than being hand-written.

Also, in my opinion, Rust is also just a nicer language to use than C - as much as I love C, I'm wildly more productive in Rust than in C/C++, and have a lot of peace of mind knowing I can more strenuously audit the behavior, security, and other aspects of the code with relative ease.

I suggest trying Rust, understanding it, and then making a judgement call. Perhaps the community is 'loud' about it for good reason.

same, I was making a remark about your comment.

also, in my and industry with much more as the stake, opinion, Ada is better.

i suggest you to try Ada.

I'm using Rust now and it seems quite nice but it could use a spec and defined semantics for provenance or whatever. It's difficult to discuss memory safety if what is actually meant is "Whatever Rust does" while the Rust docs repeatedly admit they aren't sure what Rust is meant to do in relevant areas.

Not sure what you mean. Memory safety is one thing, provenance is another. Which spec are you missing?

Well, there's no spec for any part of the language, there's a bunch of non-normative documents about how rustc might happen to work. In the non-normative documents that speculate about what guarantees rustc might require of unsafe code in order to avoid undefined behavior and ensure memory safety, provenance certainly does come up.

> why would want to write SAFE software in a language without formal spec?

A formal spec just means someone took the time and effort to write one down.

A formal declaration that a language makes no attempt at preventing memory safety is not safer than an informal declaration that a language is memory safe.

You seem to be implying that C now has a formal spec.

We have had safer alternatives for ages, Modula-2 was created in 1978, Ada in 1983, Mac OS used Object Pascal (1984),....

Your point stands though.

> We have safer options now

Not really. We just don't know the failure modes of the new hotness, and that makes us feel safer somehow. (Lol)

nghttp3 is not new software.

nghttp3 can be used to create qpack standalone binary which does not touch the network.

https://raw.githubusercontent.com/ngtcp2/nghttp3/main/exampl...

Not that it matters. OpenSSL was already 15 years old when Heartbleed was discovered, and the horrible code that caused it remained unnoticed for two years. The idea that software maturity and "enough eyes" can overcome the glaring problems from using memory-unsafe programming languages is a fantasy.

I believe the point is that this is http/3 support added to the nghttp2 library which has been around for a long time, longer than Rust's mainstream popularity started to build. While I haven't used it personally I have heard good things about the library it's good to see it continue.

Dismissing a new feature in an existing library seems quite excessive and possibly zealous. Or perhaps we should just stop using OSes entirely because they continue to release new features in memory unsafe languages.

What are you using instead of Linux?

Their comment specifically mentions that they're talking about _new_ software.

> new network-touching software

I imagine their emphasis is on the new bit here.

I’d love to have something else to use.

Just because we sometimes have to make compromises for pragmatic reasons doesn’t invalidate the original point.

ChromeOS, Android/Linux, iDevices, and Windows, where JavaScript, C++, Swift, Java/Kotlin, Rust, and .NET languages are the advised ones for new code, leaving C behind only for existing code.

"And yet you participate in society. Curious!"

Maybe it's secretly funded by the NSA.

While in reality they are coordinating with CISA to try and urgently move people away from C

https://www.cisa.gov/news-events/news/urgent-need-memory-saf...

>"It's hard for me to imagine "

Simple solution - don't

Reply on news.ycombinator.com