W3C home > Mailing lists > Public > www-style@w3.org > October 2010

Re: [CSS2.1] Clarifying 8.3.1 Collapsing Margins

From: Peter Moulder <Peter.Moulder@monash.edu>
Date: Fri, 29 Oct 2010 18:09:20 +1100
To: www-style@w3.org
Message-id: <20101029070920.GA15946@bowman.infotech.monash.edu.au>
On Wed, Oct 27, 2010 at 06:02:31PM +0200, Bert Bos wrote:

> Boxes can adjoin, elements can have siblings. To avoid issues such as 
> the one we had for "ancestor box,"

What issue is that specifically?  How was it resolved?

> let's be clear that this actually 
> means: "which might or might not be *generated by* siblings".
> But to avoid such verbose rules, I propose we continue to conflate the 
> boxes and the elements that generate them, when there is no ambiguity. 
> In particular, terms such as child and sibling refer to the tree 
> relation between the elements,

The problem is that we don't actually want "child" and "sibling" to refer
to document tree relations between document elements, nor even to relations
within the tree formed from the document element tree by suppressing
display:none elements (and certain children of display:table-column and
display:table-column-group) and adding :before and :after nodes: we want a
tree that includes the effect of 'run-in' and anonymous table object
creation and list-style-position:inside and anonymous block box processing.

The most obvious way of getting that tree is to refer to the children or
siblings of boxes, and avoid conflating boxes with elements given that the
two trees have different family relationships (i.e. the following sibling
of box A isn't in general generated by the following sibling element of the
element that generated A).

I can see that any authors who read the text would benefit from a
description based on elements, and I agree that it's an issue that we
haven't yet defined what the siblings of a box are, though I suspect
that giving an exact relation in terms of elements is at least as hard as
defining what the siblings/children etc. of a box are.

It seems quite likely that there's a misunderstanding here; have I
misunderstood Bert's suggestion?  What is the proposal for how using
element relations conveys what we want for run-in etc.: e.g. is it proposed
that we describe exceptions for run-in when talking about siblings etc., or
is it proposed that we redefine what "element" means, or something else ?

Received on Friday, 29 October 2010 07:09:51 UTC

This archive was generated by hypermail 2.4.0 : Friday, 25 March 2022 10:07:52 UTC