Aurelia Molzer

As a note while I ponder this, `::buffer` would behave kind of weirdly as returning a view containing pixels that the subiumage does not logically contain. In the mutable case...

The problem with that sample layout is its pitch information is not represented in the value and row stride as a solo method is, personally, a bit odd. We no...

The idea of returning `view` actually comes from a [much older issue](https://github.com/image-rs/image/issues/1601#issuecomment-961668855) that got take up in #2417. I admit, it was a bit of leading question. I was wondering...

You've pretty much identified the problem in `ImageDecoder` already. There's going to be a rewrite of the trait in `1.0` to take it into account. For metadata occurring after the...

Conflating threads and `rayon` together doesn't really make sense. As reminded in the linked issue there is a mode of rayon that does not create threads and also using `rayon`...

I've been drafting a bit here, but there are quite a few additional merges that happened: https://github.com/image-rs/image/tree/release-0.25.2 Anyone feel free to add notes and/or rebase on top of the current...

I think less custom structure would be better. I don't see much reason to deviate from consistency here but the type_name approach for the `debug_struct` name's parameter is pretty good...

I like the idea of unblocking data from the zlib backreference buffer by knowing that we won't use it again in a decompression call. That said the `inflater.finish` logic seems...

It's a little bit awkward for formats with more than one potential extension, choosing the first normative one is quite confusing as to _which_ hook is being called precisely. If...

Maybe we best look at other systems for handling this. Compare the web stack, i.e. everything works through explicit media types. So we'd register hooks and guesses against a media...