I’m loving the <picture> element proposal but I do see one problem with it in a practical sense rather than a functional sense: failing DRY principals.
In a real world use I would imagine (and please weigh in on this) that the adaption points for almost all images on any given page, or even any given website, would be the same points all the time. i.e., the design itself could have lets say 4 break points, and all the images are likewise likely to have 4 breakpoints. So repeating the same media rule on every single <picture> element is going to get tedious quickly. Imagine a gallery page with the same rules repeated over and over.
Can we get a way to set a “master” or default set of break-points, somewhat like CSS variables (constants), which are assumed in subsequent <picture> elements, but can be over-ridden on specific cases if needed?
Or is that out of scope for HTML?
I bring this up because I think it’s important to consider the principle of adapting content to the design rather than the device. It’s the design that adapts to the device and the content adapts to the design. In terms of prior stuff that works on this principle http://adaptive-images.com is intended to use the same breakpoints as the design in order to adapt content images.