Jon Sneyers

Well, the image/jxl images served on jpegxl.info (hosted by github.io) do not get a Content-encoding, because github.io does not seem to have an up to date MIME database. But that's...

At least same-origin HTML type declarations could in principle be trusted instead of just ignored, no? For security, I think more important than not adding new sniff patterns, is reducing...

Looks like this does indeed not support transparency. Feature request: make it work for RGBA source images as well! :) That would be very useful for tracing things like logos.

That's right. The single interleaved buffer passing is only defined for grayscale/RGB with or without alpha (so 1 to 4 channels), but if you have any other extra channels than...

Example images for viewing: Before: https://jon-cld.s3.amazonaws.com/test_images/003/jxl-e6-q80.png After: https://jon-cld.s3.amazonaws.com/test_images/003/jxl-nldc-e6-q80.png Original: https://jon-cld.s3.amazonaws.com/test_images/reference/003.png You can replace 003 by 001-049 for other images, e6 by other effort settings (e3-e8), q80 by other quality settings.

I don't think we should ever ignore an explicit -d/-q, but I do think it makes sense to have a different default -d/-q depending on how the input is given...

Since the png looks black, I assume you only use the low part of the range — e.g. the actual data is 10-bit, and you encode white as 1023 instead...

At the moment I don't think we have an API that can take 'uint11' input or produce such output. Maybe we should change that, because it would indeed be more...

> From my perspective, where I'm integrating jxl as a codec inside libtiff, what would simplify my life more than high vs low bit padding, would be to support a...

> Discussing this again gave the following conclusion (in contrast to the post just above): > > * Having `SetPreferredColorEncoding` and `JxlDecoderSetDesiredIntensityTarget` not working in all cases is a problem...