Re: [fxtf-drafts] Clarify the size of the "input image" (#324)

The recent changes for the `backdrop-filter` definition require the opposite of Gecko's approach.

It seems to me that the intention of "wrap" and "duplicate" is to prevent fading at the edges of the image. But here's what you actually get.

![edgemde](https://user-images.githubusercontent.com/989243/63965651-8ad82300-ca91-11e9-961e-bebd9f923395.png)

Those are the results of this input:
```xml
<filter id="filter">
  <feGaussianBlur stdDeviation="8" edgeMode="nnn" />
</filter>
<g style="filter:url(#filter)">
  <rect style="fill:red" x="0" y="0" width="50" height="25"/>
  <rect style="fill:blue" x="0" y="25" width="50" height="25"/>
</g>
```

The image has a 5 pixel transparent boundary around it. The two images with the (*) are the result if you wrap the "input pixel" to the input bounds (0, 0, 50, 50) instead of the filter bounds (-5, -5, 60, 60)

The main things to note:

* If you wrap at the filter-bounds, the "none" and "duplicate" images are identical - because the pixels at the edge of the image are always transparent.
* Although the change in hue shows the color "wrap", there is also a white fringe where the transparent pixels have been blurred.

To me, it looks like wrapping at the edge of the input image, as required by `backdrop-filter`, makes more sense.

-- 
GitHub Notification of comment by faceless2
Please view or discuss this issue at https://github.com/w3c/fxtf-drafts/issues/324#issuecomment-526306858 using your GitHub account

Received on Thursday, 29 August 2019 18:29:05 UTC