- From: Simon Fraser <smfr@me.com>
- Date: Fri, 11 Jun 2010 15:44:27 -0700
- To: fantasai <fantasai.lists@inkedblade.net>
- Cc: Brad Kemper <brad.kemper@gmail.com>, "L. David Baron" <dbaron@dbaron.org>, Brian Manthos <brianman@microsoft.com>, "www-style@w3.org" <www-style@w3.org>
On Jun 11, 2010, at 3:18 PM, fantasai wrote: > On 06/11/2010 02:03 PM, Brad Kemper wrote: >> On Jun 11, 2010, at 11:13 AM, Simon Fraser <smfr@me.com> wrote: >> >>> I don't think the current definition, which describes the blur in >>> terms of a gradient, is good for shapes with concave portions. >> >> I don't know why not. It doesn't say it's a gradient, it just defines >> the size of the region to blur within. I think that saying that a 15px >> blur covers a perimeter that is 15px wide will be a whole lot more >> understandable and predictable and meaningful for authors than to ask >> them to guess how much that will be based on the results of plugging >> that length into a guassian function. > > What Simon is trying to say is that it's not a straight-up transition > of 15px. > > If I'm understanding this correctly (I'm shooting in the dark here), > the Gaussian function, when applied to concave shapes like the inside > of a corner, will result in an effective "radius" that is much larger > at certain points. This is in fact what you want: otherwise the corner > doesn't look blurred, it looks gradient-ed. > > Imagine a sharp concave corner (i.e. the border with an inner shadow). > If you put a true Gaussian blur on that, the edge where the shadow > finally disappears will have a slight curve. > > +-------------- [ I lack hixie's awesome ascii art skillz, > | but I'm trying here... ] > | _____ > | ,' > | : > | | > | | > > In the current definition, you'll get a sharp edge. > > +-------------- > | > | ________ > | | > | | > | | > | | > > If I'm understanding this correctly, applying a true Gaussian and > then thresholding it will probably fix those weird kinks you were > seeing on inner shadow spreads with the current definition. Yes, this is exactly the issue. Simon
Received on Friday, 11 June 2010 22:45:21 UTC