[PNG] Meeting topics & info

Hello everyone,
We have a meeting this coming Monday. Here are some updates we can talk
about:

   - 4th Edition:
      - SMPTE 2094-50 <https://github.com/SMPTE/st2094-50> (gain curves)
         - An updated Public Committee Draft (#2) was just released earlier
         this week.
         - I believe we're waiting for AOM to make an equivalent. Do we
         have any AOM members? Do we know the status of the equivalent?
      - ISO 21496-1 <https://www.iso.org/standard/86775.html> (gain maps)
         - Any updates here on how we should add this?
      - 5th Edition:
      - Test images
         - I've gathered several public domain images to use as a test. I
         tried to pick ones I believed represent the type of image
someone would
         save as a PNG. I am temporarily hosting them on my website
         <https://www.programmax.net/PNG_test_images.zip>. Eventually, I'll
         make a proper page so finding them is easier. When that
happens, I will
         break that link.
            - Many of the images are SVG, which means you would need to
            render them at some given canvas size and quality. To keep
things simple, I
            just used https://svgtopng.com/ . The way it chooses canvas
            size is not clear, but it also isn't terribly important
for generating a
            first batch of test images.
         - I computed the SSIM of the first image convertex to a JPEG XL
         using -quality 100 and was given a SSIM score of 0.442165. That is a
         surprisingly bad score--especially given it was 100% quality
target. The
         JXL itself looks good. It makes me wonder if SSIM is our best bet. In
         either case: 1.) I'll make a full report and send it later,
and 2.) this
         very early indicator suggests there is plenty of room for a
better lossy
         image format for logos and icons.
      - Compressor updates
         - I was given a suggestion to look at ECT
         <https://github.com/fhanau/Efficient-Compression-Tool> and BZip3
         <https://github.com/iczelia/bzip3> when researching compression
         alternatives. ECT is backwards compatible and seems to perform well.
            - I used ECT on an image from the test images above and it
            compressed a PNG to be smaller than a lossless JPEG XL
(default settings on
            all tooling). I'll look into this further, but that's
already significant.
            - The impressive performance of ECT also raises the bar for
            alternative compressors like zstd. Will alternate
compressors still be able
            to beat old DEFLATE? If so, is the delta still significant
enough to
            warrant a new compressor?