- From: Silvia Pfeiffer <silviapfeiffer1@gmail.com>
- Date: Thu, 7 Apr 2011 10:56:18 +1000
- To: Philip Jägenstedt <philipj@opera.com>
- Cc: public-media-fragment@w3.org
On Thu, Apr 7, 2011 at 1:23 AM, Philip Jägenstedt <philipj@opera.com> wrote: > On Wed, 06 Apr 2011 15:05:43 +0200, Silvia Pfeiffer > <silviapfeiffer1@gmail.com> wrote: > >> Yeah, I think we need to be more specific. My suggestion is to round >> up, but am curious what ppl think. > > Do you mean that if the original rect (x, y, w, h) is (10.4, 20.5, 30.5, > 40.6), that all values are rounded up to give (11, 21, 31, 41)? That would > cause a whole row and colums of pixels to the right and bottom to be > included that might not have been intended. > > To avoid adding up rounding errors from x+w or y+h, I would prefer > transforming the (x, y, w, h) to a (left, top, right, bottom) representation > (10.4, 20.5, 40.9, 61.1) and work on that. I would prefer one of: > > 1. rounding all values, giving (10, 21, 41, 61) -- closest fit > > 2. rounding "outwards", giving (10, 20, 41, 62) -- always includes the > entire targeted region That's what I meant with "rounding up". Jack explained it perfectly. Silvia. > > 3. rounding "inwards", giving (11, 21, 40, 61) -- never includes anything > outside the targetet region > > Some consistency with how CSS works would be great here, but I don't know > how CSS works. > > -- > Philip Jägenstedt > Core Developer > Opera Software > >
Received on Thursday, 7 April 2011 00:57:06 UTC