Re: [css3-multicol] unknown available width and shrink-to-fit

Le 17/10/2012 17:06, Håkon Wium Lie a écrit :
> This would make the pseudo-algorithm simpler. Simpler is generally
> better. But if we just move complexity to somewhere else (in time or
> space), we may noe gain much. If we remove lines 03-10 and replace
> 'available-width' with 'used-width', the input to the pseudo-algorithm
> would change: 'used-width' would have to be known. What should the
> spec say about finding it -- just point to shrink-to-fit in 2.1?

I think that CSS 2.1 does a pretty good job at separating how a box 
behave "on the inside" and "on the outside".

For example, block-level and table cell are two "outside" behaviors; 
block container and replaced box are two "inside" behaviors. All four 
combinations are possible.

Multicol is only a new "inside" behavior, it does not affect the 
outside. I think it shouldn’t (directly) affect how §10.3 of css21 
determines the used width.

Some cases in §10.3 use the "shrink-to-fit" width, which css3-sizing 
terms is defined as:

   fit-content = min(max-content, max(min-content, fill-available))

Now, if a box happens to be multicol, this changes its max-content and 
min-content widths but not the above definition of fit-content or the 
fact that the used width of a float with 'width: auto' is fit-content.

The multicol algorithm determines the used column-width and column-count 
during layout, based on a known used width. Determining the intrinsic 
max-content and min-content without layout is an entirely different job.

Simon Sapin

Received on Wednesday, 17 October 2012 15:44:32 UTC