Timothy Lochner

I think 3 also creates a problem where the rand across threads is not guaranteed to be decorrelated but I guess crt rand is very poor anyways so it doesn't matter. I kind-of like it actually, because it's so poor you (should) know it's just a "casual" rand. Writing a "good" one..

...is really context dependent, so it's probably best left to the application developer. C++ tried recently to have a "good" rand and it ended up to be a gigantic ugly and still flawed thing.

This is a general problem with the C++ std lib IMO. They try to build Perfect Solutions that will cover 100% of use cases. Better to make the 99% use cases easy/intuitive. The 1% folks are always going to use their own libs anyway, so why make the 99% suffer to cater to them?

So right! The idea that a perfect solution to all use cases exists is absurd anyway. A math library for scientific needs is too costly for game dev, but a game dev math lib is too inaccurate for scientific needs. They have conflicting needs.

I'm becoming of the mind that the std lib should only provide abstractions users cannot create themselves, that the OS has to provide. Thread, filesystem, allocating memory, etc. Everything else should be open source libs.