Don't set intrinsic width from sizes since that's bogus. Don't set the i...

[zcorpan]

...ntrinsic resolution here; we return the density and the HTML spec handles the resolution with 'current pixel density'. Fixes #121 Fixes #107

(We may want to return the w/h descriptors too as part of #85.)

[yoavweiss]

Can you add a note saying that intrinsic sizing is handled by HTML, maybe with a link to where that happens?

[tabatkins]

Where does intrinsic width get set, if not here? In HTML?

[zcorpan]

"The user agent is expected to treat the element as a replaced element and render the image according to the rules for doing so defined in CSS."
http://www.whatwg.org/specs/web-apps/current-work/multipage/rendering.html#images-0

"For raster images without reliable resolution information, a size of 1 px unit per image source pixel must be assumed."
http://www.w3.org/TR/CSS21/conform.html#intrinsic

http://www.whatwg.org/specs/web-apps/current-work/multipage/edits.html#current-pixel-density overrides the density.

Nothing sets the intrinsic dimensions before the image is loaded, but we could do that as part of #85.

[tabatkins]

Okay, my setting of the intrinsic width here in the algo was intentional. That's part of the point of sizes=''.

[zcorpan]

I thought the point of sizes was to give the UA a ballpark estimate on how big the image is going to be in the layout, so that it can download an appropriately sized image from the list. I fail to understand how such a value is appropriate for the intrinsic width. The width descriptor divided by the density would be a more accurate source for use as the intrinsic width, but it still seems useless to set intrinsic width without also setting intrinsic height.

[tabatkins]

Oh wait, on further consideration, you might be right. It's hard to remember precisely what my earlier intention was. I suppose the intrinsic width is best set from from w descriptor.

I agree that having an intrinsic width isn't greatly useful without an intrinsic height, which is why I planned to allow h descriptors as well. We haven't added those yet, though. (Tracked in #86.)

[zcorpan]

So OK to merge?