Aurelia Molzer

Huh, I wasn't aware `num-traits` was actually implementing some IEEE math itself without `libm`. My perception was they'd just defer to one of the available methods.

After some consideration of the API, there's some good ideas for a renovation and `2.0` release anyways. For instance [`map_pixel`](https://docs.rs/color_quant/latest/color_quant/struct.NeuQuant.html#method.map_pixel) and `index_of` should take arrays respectively. Thus, I'd personally be...

> Also, do you have a better idea for the feature's name because alloc is not really appropriate. Maybe num_traits? Something else? Generally, features should be (super-) additive, not subtractive....

So, there could be merely two features: `std`, and `num_traits` as an optional dependency. Then the crate would require _either_ to work, and ignore `num_traits` if `std` is also selected.

I'd prefer `1.56` and adding [`[package.rust-version]`](https://doc.rust-lang.org/cargo/reference/manifest.html#the-rust-version-field).

> J40 needs to use many image buffers; should we use [ImageBuffer](https://docs.rs/image/*/image/struct.ImageBuffer.html) for this purpose, or make our own? (This is important when J40 is going to be used as...

For unsupported color layouts, #2522 should be the general direction. It's probably _fine_ to convert the color while encoding. At least truncating the depth was a sought after compatibility adjustment...

Sorry, I forgot to fill out additional context: It's mentioned in `uutils/coreutils`:

My grains of salt, using SeqCst is a misinterpretation of the kernel documentation: > When using the SQ poll thread (IORING_SETUP_SQPOLL), the application needs to check the SQ flags for...

After having another think about this and re-re-reading the diagrams in the original issue, I don't think I really understood the cross-language-atomic-semantics properly yet. That might mean there are not...