Re: [CSS21] Question about width calculation for replaced elements with intrinsic percentage width

CSS determines the width of boxes in one of two ways: outside-in or 
inside-out. The former is for boxes in normal flow: the sum of width 
and margins is equal to the width of the containing block, the width of 
child boxes doesn't play a role. The latter concerns boxes that 
shrink-wrap: the width depends on child boxes, and the containing 
block's width only provides a hint in case there are multiple possible 

Of course, as Boris points out, a box in normal flow may have a 
containing block that itself is shrink-wrapped, and in that case the 
box's children may influence the box's containing block, and thus the 
box is *indirectly* influenced by its children. But that is not what is 
meant in 10.3.2 (nor anywhere else where the same phrase is used).

The phrase ("if the containing block's width depends on this element's 
width") was used instead of an enumeration of cases, because we expect 
there will be more cases in CSS3 than in CSS2. In particular, there 
will be the case of a block-level box in normal flow that is 
nevertheless shrink-wrapped, because it has 'width: fit-content'. (In 
practice, though, the sections will be rewritten anyway, because of 
those extra values and because of the addition of vertical text.)

But with respect to level 2, the phrase "if the containing block's width 
depends on this element's width" refers to containing blocks that are 
made from:

  - Floating, non-replaced elements
  - Absolutely positioned, non-replaced elements
  - 'Inline-block', non-replaced elements in normal flow

On Monday 04 February 2008 20:48, fantasai wrote:
> Boris Zbarsky wrote:
> > fantasai wrote:
> >> Ok, so my understanding is that for CSS2.1:10.3.2, we should say
> >> that if a
> >> replaced element has intrinsic percentage width and its containing
> >> block's
> >> width depends on the replaced element's width/height, then "the
> >> resulting layout is undefined in CSS2.1", as it is for specified
> >> percentage widths in the same situation.
> >
> > I guess so, yeah.
> Exact replacement:
> In Section 10.3.2, last paragraph replace
>    "If it does, then a percentage intrinsic width on that element
> can't be resolved and the element is assumed to have no intrinsic
> width."
> with
>    "If it does, then the resulting layout is undefined in CSS2.1."

That's fine for CSS 2.1.

We can define it precisely in level 3, I think. The CSS box module 
level 3 has text for how to determine the width of shrink-wrapped 
boxes. (There are still some paragraphs marked as "to do", but we'll 
get there...). That text also handles descendants with percentage 

> >> For 10.6.2 we change nothing, except move the paragraph beginning
> >> "Percentage" up to become the second paragraph to clarify its
> >> interaction with the other rules in that section.
> >
> > Yeah, that would make sense.
> ~fantasai

  Bert Bos                                ( W 3 C )                               W3C/ERCIM                             2004 Rt des Lucioles / BP 93
  +33 (0)4 92 38 76 92            06902 Sophia Antipolis Cedex, France

Received on Tuesday, 5 February 2008 14:33:16 UTC