Aurelia Molzer

Not without getting rid of `impl Pixel for Rgba` since this impl overlaps those individual ones, which is forbidden.

> It should be a backwards-compatible change, except for the generic parameter now being replaced with concrete primitive types. I'm not a fan of taking chances on this one. It...

That's what I was trying to say: Definitely enough to justify going to the next major version, not enough to disregard SemVer and try to come up with justifications to...

> And AFAIK for many decoders this isn't actually true at all: they try to bail with an error as soon as they discover that any pixel data is corrupt,...

:+1: on deprecation from codecs when a crate is ready but I'd be slightly more cautious on removal. It would be great if we could change the implementation without directly...

The only case where this is _less_ general is when `F` is recursive, right? We never required `Send` or `Sync` so it's not possible to call the callback from a...

Well, not sure if that was the right button. My brain assumed it was the `Approve for Run` for reasons of placement… It seems like it wasn't completely wrong though...

I've create a new branch `next-version-0.25` for all breaking changes. Feel free to amend everything into the breaking change then.

That comment is no longer accurate [since 0.7.0](https://github.com/image-rs/image-tiff/blob/master/CHANGES.md#version-070), although tiled tiffs need to be supported manually: [`Decoder::read_tile`](https://docs.rs/tiff/0.7.1/tiff/decoder/struct.Decoder.html#method.read_tile). Manually meaning that there is no support for selectively retrieving tiles with tile...

Honest, I'm just delaying here because I don't get it either :slightly_frowning_face: It all seems like action at a distance how all the variables work together.