- From: Tab Atkins Jr. <jackalmage@gmail.com>
- Date: Tue, 15 Oct 2013 11:56:29 -0700
- To: "Kang-Hao (Kenny) Lu" <kanghaol@oupeng.com>
- Cc: Morten Stenshorne <mstensho@opera.com>, WWW Style <www-style@w3.org>
On Tue, Oct 15, 2013 at 11:13 AM, Kang-Hao (Kenny) Lu <kanghaol@oupeng.com> wrote: > (2013/10/16 1:46), Tab Atkins Jr. wrote: >> On Tue, Oct 15, 2013 at 10:43 AM, Morten Stenshorne <mstensho@opera.com> wrote: >>> "Tab Atkins Jr." <jackalmage@gmail.com> writes: >>> Are there similar problems in other specs, and are there any solutions >>> out there already that can be of inspiration? >> >> Yes, there are a number of places where the natural range for some >> value is open, and we've patched it into being closed. If the error >> preventing me from editting the wiki is fixed, I'll have a nice >> write-up with these solutions. >> >> The three solutions we've used are: >> >> 1. define a minimum size, and clamp things that are within the >> boundary but less than the minimum. > > Which CSS property does this? I'm not sure if any do this yet, but it's a reasonable behavior. >> 2. find the limit behavior, and say that below a ua-specific limit, >> it's just that limit behavior. (Repeating gradients do this.) >> 3. define a completely different behavior for the boundary value >> itself. (background-size/repeat do this.) > > 2. makes the most sense to me and I wonder why we didn't do the average > color thing for 'background-size: 0'. Probably because it seems expensive, but mostly because we didn't have this rule established as firmly in the past. > I am inclined to think we should do 2. here too, even if it might be > arbitrarily slow (is that the concern here?). > > Or we can as well just say 'column-width: 0' has to be valid but the > behaivor is implementation-defined. This seems worse than defining a reasonable behavior. > Authors should be able to do > whatever he/she wants in 1px x 1px and scale it up with CSS transform > without seeing quantum effects. While an admirable goal, rounding and precision issues guarantee that this won't be true - the browser environment is inherently quantum (in at least two cases, quantized to 1/60 of a px). Due to this, I don't feel terrible strongly about trying to do anything that would keep things precise *if* browsers had infinite precision. ~TJ
Received on Tuesday, 15 October 2013 18:57:17 UTC