all-fakes

Yeah, that's the obvious counterexample, but that's why I said "more often than not". The statement I was questioning was that fast-math is "normally not more correct."

I thought fast-math meant (among other things) that the compiler could algebraically simplify things using some rules for real numbers that don't apply to 32-/64-bit floating point numbers. I'd expect that reducing the number of operations would more often than not reduce the error due to rounding, which is what most people would want when they talk about "correctness".

Which step in this (very naive) argument is wrong?

> In the same way your stance on climate change is a scientific position regardless if your stance is correct or not.

Can you please just give the definition of "scientific position" that you are using?

Do you mean an assertion regarding reality? Because I think everyone agrees on the reality. The question of what the words "sex" and "gender" constitute is not a question of reality: it is a question of social and legal constructs.

> It's still an evil to unjustly attack someone even if to different degrees.

Yep.

Look, I agree. I shouldn't be forced to do something for someone who makes me uncomfortable, even if it's something that they can't change, and which doesn't materially impact me in any way.

But at the same time, do you not see how this reinforces existing disparities in opportunity and social acceptance? If you have a better solution than mandating that businesses don't discriminate, I'm all ears. However, if you simply allow businesses to freely discriminate, then the world will only become more unfair over time.

> It's a scientific and political position.

What exactly do you mean by "scientific position"? The term seems, to me, an absurdity.

> Doesn't make it okay.

Putting aside any actual wrongdoing, attacking those with privilege is infinitely better than attacking those without: the former have the resources to defend themselves; the latter do not.

Sorry, can you explain why the velocities should add that way? I've always just seen velocities add like regular old vectors, so this looks very strange.

That is easily avoided by having them sign off on the code they commit to the employer's repository. Giving the employer full rights over every single thing the employee creates just to solve such a simple problem is ridiculous.

> And yet intuitively the function is increasing and so we ought to define "having a derivative" in such a way that shows that it is increasing.

The Cantor function is not differentiable at the points of "increase", so there is no problem with it growing at those points. I'm not sure what your complaint is.

> it seems like a bunch of silly definitions with silly consequences, with little relation to the actual mathematics of our universe that I care about.

That's honestly really silly. A lot of analysis is extremely practical, and is fundamental to the study of this universe. Yes, the language allows you to express non-physical situations, but you know you can just ignore those situations, right?

Yes. Some libraries, like typed-arena (and Rust's built-in Vec), will traverse the structure and call the drop implementations, and others, like bumpalo, will just leak the resources.

Typed-arena is a bit limited, because all the allocated values must be of the same type. bumpalo[0] removes that restriction, but it also doesn't run destructors, which can lead to memory and other resource leaks. Another worry I have with bumpalo is that I don't think it's been reviewed thoroughly for unsoundness.

Still, bumpalo has been used to great effect in dodrio[1], a React-like library for Rust with really good performance.

[0]: https://crates.io/crates/bumpalo

[1]: https://github.com/fitzgen/dodrio

Yes, Java also has a garbage collector that does nothing, called the Epsilon GC, intended for short-lived programs and references for garbage collector benchmarks.[0]

[0]: https://blogs.oracle.com/javamagazine/epsilon-the-jdks-do-no...

I'm not sure what the performance overhead would be, but that state can easily be stored on disk.

If we're being pedantic,

> Some bundlers may even have O(n^2) complexity: as your project grows, your dev environment gets exponentially slower

that's clearly not exponential!

What views are you insinuating that I have on "election interference using social media in 2016"?

How do you know that they aren't just as critical of their own government?

> why? who cares? do you not know how to skim and skip irrelevant text?

Ignoring inflammatory content can be more taxing to people than you pretend, and moderators and flagging are too slow to act. It's not the end of the world, but it's annoying enough that I would consider alternative services.

A forum is nothing without its users, and a forum that puts its users first shouldn't irritate its users with off-putting content without good reason.

> you could think of any number of ways to vet comments.

Alright, I was just checking whether you had any new ideas, and it seems that you do not.

> you're wrong that there are fewer submissions to arxiv

HN receives at least twice as many submissions per month, and an order of magnitude more comments. This makes the arXiv a very poor analogy for HN.

> you're also wrong that it's moderated

From [0] (see also, [1]):

> All submissions are subject to a moderation process that verifies material is appropriate and topical. Material that contains offensive language, non-scientific content, or is plagiarized may be removed.

Looks a lot like they have moderation to me.

> i also don't understand how heavy moderation is a counterpoint

If you already heavily filter by quality, then the order in which posts are presented obviously becomes much less important.

> yc is one of the most successful vcs in the world - they can't afford moderators?

HN does have an excellent moderation team. They just aren't anywhere near as stringent as the arXiv, on quality, on politeness, or on any number of other characteristics.

If HN was as heavily moderated as the arXiv, then they wouldn't need a voting system either. It'd be a much colder place, though, which is probably why they don't do that.

[0]: https://arxiv.org/help/submit

[1]: https://arxiv.org/help/moderation

> This presupposes some kind of universal value function.

Sorry, I did not mean to suggest that some comments are intrinsically more valuable. But we do want to sort comments so that, on average, the distance between the "ideal" order for any given reader and the actual order is minimized, don't we?

> This is like perfect is the enemy of good counterargument.

I'm not trying to argue anything here, I just wanted your thoughts on how this could be implemented effectively.

> I have never browsed a journal or arxiv by the number of citations or the hindex or whatever.

I don't know if arXiv is a very good comparison — the posts are much fewer, the range of interests is much narrower, and the site itself is very heavily moderated.

Responses have to be ranked, because some comments are much more valuable to the reader than others. Well-thought-out comments and spam lie at the extreme ends of the spectrum.

I think there's merit to the idea of requiring a response for upvotes or downvotes, but how can you tell whether a response is "genuine"? It sounds like it'd devolve into people leaving vacuous replies just to clear that prerequisite.