Re: [css3-intrinsic-sizing] sizing percentage width content in quirks mode

On Wed, Aug 8, 2012 at 3:03 PM, L. David Baron <dbaron@dbaron.org> wrote:

> On Wednesday 2012-08-08 14:25 -0700, Ojan Vafai wrote:
> > Before I start fixing bugs, what should we do in these cases? It seems to
> > me that respecting min-content/max-content is a no-brainer.
>
> I presume you're talking about respecting
> width:min-content/max-content on an element E when computing one of
> the intrinsic widths of E's parent P.
>

Correct.

Here's a better example of case 3 where WebKit and Gecko disagree:
<div style="width: 100px; background-color: salmon">
<div style="display:inline-block; border: 5px solid orange; width:
min-content;">
    <div style="display:inline-block; width: max-content;">abcdefg
hijklmndop</div>
</div>
</div>

The Gecko behavior is clearly better. I'd just like to see it clearly
specced. I'll file a bug to fix this in WebKit.


> > Should we also
> > respect percentage widths, fill-available and fit-content which depend on
> > the available width? It's certainly what I would expect as a web
> developer.
> > I'm not sure if we can safely change percentages without breaking
> existing
> > sites though.
>
> I don't think respecting 'fill-available' and 'fit-content' requires
> any change; the correct behavior is, I think, exactly the same as
> for 'auto', for both intrinsic widths.


Seems reasonable.


>  (The correct behavior of a
> preferred intrinsic width is either (a) to have no optional breaks
> or (b) to take up as much width as can be usefully occupied.  The
> correct behavior of a minimum intrinsic width is to find the
> smallest width that doesn't lead to overflow that could be avoided
> by choosing a larger width.)
>
> For percentages, we could honor percentages by multiplying the
> relevant intrinsic width of E by the reciprocal of the percentage to
> find its contribution to the intrinsic width of the parent.  This
> would produce the correct behavior as described above, but I don't
> think it's Web-compatible with respect to the 'width' property,
> outside of a table context (where percent widths on cells do,
> interoperably, behave this way).  Gecko does do this behavior for
> the preferred intrinsic width, for percentage margin and padding
> (where the formula is different, since it requires multiplying the
> non-percentage part by the reciprocal of 1 minus the sum of the
> percents).
>

Maybe the best thing to do for percentages is to treat intrinsic sizes the
same way we do auto as that adds the least complexity and does something
reasonable-ish?


>
> -David
>
> --
> 𝄞   L. David Baron                         http://dbaron.org/   𝄂
> 𝄢   Mozilla                           http://www.mozilla.org/   𝄂
>