W3C home > Mailing lists > Public > www-style@w3.org > February 2012

Re: [css3-regions] regions forming stacking contexts

From: Anton Prowse <prowse@moonhenge.net>
Date: Fri, 03 Feb 2012 11:42:59 +0100
Message-ID: <4F2BBA33.1020907@moonhenge.net>
To: www-style@w3.org
CC: Alex Mogilevsky <alexmog@microsoft.com>
On 01/02/2012 00:12, Alex Mogilevsky wrote:
> From: Vincent Hardy [mailto:vhardy@adobe.com]  Sent: Tuesday, January 31, 2012 4:36 AM
>> On Dec 27, 2011, at 1:58 AM, fantasai wrote:
>>> # Regions create a new stacking context. [...]
>>> # With regions, an element may be split across multiple regions
>>> # and these regions may overlap (for example if they are absolutely
>>> # positioned). So fragments of the same element can overlap each
>>> # other. Since each element has a single z-index, it would be
>>> # required to find another mechanism to decide in which order the
>>> # fragments are  rendered. Since each region creates a new stacking
>>> # context, it is clear that each region is rendered separately and
>>> # their rendering order follows the regular CSS rendering model.
>>> Fragments of the same element can overlap each other already due
>>> to inline breaking, so this situation is not unique to regions.
>>> Unless there's some compelling reason why they should be different,
>>> I think I'd rather regions "followed the regular CSS rendering
>>> model" as you say at the end... and did not create a new stacking
>>> context unless other CSS properties on the region dictated it.

>> Our team believes that since we see regions as a low level building
>> block, which, for example, could be used (internally) to implement
>> multi-col, indeed, the rendering model should be no different and we
>> could/should make the change you suggest.
>> Alex thinks differently but I'll let him make his argument.

> I thought about it. It seems both ways should be possible, and the
> argument about regions being low-level building blocks is good.
> It will be easier for me to agree if I see an example where regions
> render atomically (as would be expected when regions are used as
> pages). Is it possible to achieve with existing options? Will it be
> enough to make regions 'position:relative' (I don't think so, just
> checking)?

Using properties that induce stacking contexts (such as opacity, or 
relpos with integer z-index) on an element which has a non-'none' value 
of 'flow-from' (ie, on a region) ought to be enough to make the region 
establish a stacking context.  Is there a reason to break that 
established behaviour?

(Of course, whether it's practical for some regions to be stacking 
contexts and others not is a separate question....)

Anton Prowse
Received on Friday, 3 February 2012 10:43:27 UTC

This archive was generated by hypermail 2.3.1 : Monday, 2 May 2016 14:38:55 UTC