I've filed an issue with SVG, as displaying SVG in an image doesn't seem well-defined w3c/svgwg#289
However, we need to define how SVG is drawn to a canvas, particularly in terms of media queries.
Edge uses the layout size of the <img> to determine the viewport. If the <img> doesn't have layout then it falls back to the width/height attributes, if it doesn't have those it falls to the intrinsic dimensions of the <img>. This seems ideal, as you can change the viewport independent of the size it's drawn to the canvas.
@junov@xidachen@grorg any opinions on this question of drawing a SVG to a canvas? Does aligning with Edge's behavior here sound reasonable, if we write some web platform tests and update the spec to be precise?
I agree with @jakearchibald that Edge's behavior is probably the more desirable one for web devs. The "bizarro-pants" behavior in Chrome is based on a (possibly incorrect) literal interpretation of the spec, which I agree is not that useful.
Implementations are not interoperable with each other
Standard is not web compatible or proprietary feature needs standardizing
Mar 28, 2017