>> most binary search implementations don't claim they only work under some particular inputs
> They do implicitly. It's just common sense.
That's neither how language specifications work, nor true in this case even if it's true in other cases. Providing one more of the same kind of input that already works is in no way the same thing as changing something totally unrelated.
> When you read a recipe in a cookbook, it usually doesn't mention that you're expected to be standing on your legs, not on your arms.
I don't think this binary search was breaking because of people standing on their arms either.
> A lot of generic algorithm implementations will start acting weird if your input size has the order of INT_MAX. Instances this big will take days or weeks or process on commodity CPUs,
It's incredibly strange to read this from someone in 2022. I don't know of any standard library algorithm that would take "days or weeks" for inputs of size 2^31 now, let alone the majority of them being like this. In fact I don't think this was the case back when the article was written either.
