Jonathan Behrens

This PR is far too large, with for too much all rolled up into it to meaningfully review. The proper way to add a big feature would be starting with...

I don't fully recall the state of that other PR, but it is more than 5x smaller than this one. TBH I haven't figured out how to handle multi-thousand line...

Do you mean the "Created global ifd definition" change? Haven't looked in detail, but from skimming: * **Making `Decoder` object to be generic over normal vs bigtiff:** I don't think...

What file format are your images stored in? While this crate makes simplifying assumptions to abstract over all the formats it supports, the backend decoding crates generally expose more features...

That interface seems fine to me, but I'd also be OK with something simpler like just having a `sbit: Option` field on `Info` and adding note that any elements beyond...

Looks good. We'll just want to hold off merging until we've gotten out a minor release with all the non-breaking changes we want to land

So there's a couple different aspects that you might want to abstract over: * Being generic over what `Pixel` trait implementation you're dealing with. So being able to abstract over...

> I think we should focus on making it easy to write code that covers all variants of DynamicImage first and foremost. Being more generic than that is a nice...

The `ImageBuffer` type (and by extension `DynamicImage`) currently represents pixel data simply as `Vec`, `Vec`, `Vec`, etc. These can be cast to byte slices or `&[[T; N]]` using bytemuck or...

To be clear, `ImageBuffer` has row continuous views already. In fact, it goes even further and gives you a view with all rows packed together into a single slice. That...