core(modern-image-formats): ignore images CF claims are bigger as webp

[connorjclark]

closes #14265

#14265 (comment) points out that while hand-tuning could really result in some savings, it seems OK to defer to CF for users of their automatic conversion feature.

[brendankenny]

One source of confusion for users is that modern-image-formats is actually estimating savings when converting to AVIF since #12682 (in the 8.1.0 release), but the description still includes "WebP" (in fact, it puts "WebP" before "AVIF").

We discussed in #12682 (review) the original plan of moving the WebP estimate to uses-optimized-images, but decided not to and just leave the WebP estimate in the JSON, but we left WebP in the description. Maybe we just need a description update? That means we don't have a recommendation for WebP anywhere, though...

[connorjclark]

couple things we may want to do here (not mutually exclusive)

(1) drop WEBP from description
(2) add "webp savings" column (and denote the existing savings column as "avif savings")
(3) do an actual conversion in OptimizedImages (needs protocol support)

---

(1) in isolation means we no longer mention webp in the report anywhere
(2) in isolation means we just duplicate information, since one value is computed from the other
(2) + (3) seems good but....

(3) is a lot of work.

https://source.chromium.org/chromium/chromium/src/+/main:third_party/skia/src/images/SkImageEncoder.cpp;l=58;drc=a442c20f87b3f20ec085cc20f86938f82b8c6914 shows all the formats skia supports for encoding

there is of course a avif codec in skia, but it only decodes https://source.chromium.org/chromium/chromium/src/+/main:third_party/skia/src/codec/SkAvifCodec.cpp;l=1?q=src%2Fcodec%2FSkAvifCodec.cpp&ss=chromium%2Fchromium%2Fsrc:third_party%2Fskia%2F

skia uses libavif which does support encoding https://github.com/AOMediaCodec/libavif so it'd be a matter of tapping into that, then exposing this to the protocol in the same way the other formats are.

should we pursue (3)?

---

back to the PR... it seems that we can just close this given we aren't even comparing the same thing (avif savings vs CF's webp)

[connorjclark]

closing for #14592