Granted, even at 11 kb the image looks really good and there are no visible artifacts but when you put the original image and the lossy version side-by-side, you can see that some tones are slightly off and gradients/shades are slightly different.
Granted, even at 11 kb the image looks really good and there are no visible artifacts but when you put the original image and the lossy version side-by-side, you can see that some tones are slightly off and gradients/shades are slightly different.
The same image can be made into a 13kb PNG (8-bit + imageoptim), without any complex trickery and 99.5% browser support: http://f.cl.ly/items/303H1k2c412s1o0m1b2m/webp-8.png
cwebp -lossless -pass 10 -m 6 MaxCDN-WebP.png -o image.webp
Still that's not a bad compression improvement given that 'pngcrush -brute' failed to improve compression and pngout managed only to bring it down to 50.1kb
With JNG those massive parallax websites could be ~1MB instead of ~5MB.
But lossy + alpha was a good idea then and it still is now.
In case you didn't know, there are 3 very important browsers: Chrome, Firefox, and IE. Right now, Firefox got around 20%, IE around 30%, and Chrome around 40%.
There are also a few countries where Firefox is the most popular one. For example, in Germany it currently got around 45%.
>a fringe image format
Was PNG any different? People usually don't randomly start to use some image format if it isn't supported by any browser.
Also, ripping two images out of one file and splicing them together requires very little code. In a high-level language it's maybe 50 lines of code. There are many games which do that. It's very simple and it saves typically around 80%.
Interesting to see Hixie in there advocating for MNG (or, rather, for a format with 8-bit alpha and animation that's not Flash).