Re: [css-flexbox] Behaviour of percentage heights in column direction

On 02/24/2015 03:52 PM, Greg Whitworth wrote:
> There are so many similar threads (and some splitting threads)
> that have impacts on the other threads recently raised by
> Daniel

Yeah -- there are 3 distinct issues that I've raised recently, and they're orthogonal, though all tied into content-measurement stuff.

For reference, here are summaries of the three issues, w/ links to the threads on the w3.org web archive. (Of course, for sanity's sake, any thoughts/responses about these issues should go to the corresponding thread; I'm just listing them here to distinguish the issues.)

 (1) This thread here: "should we make 'min-height:auto' taint the resulting flexed height on a vertical flex item & make it indefinite?" (I slightly lean towards 'yes' for perf reasons; Greg disagrees; I'm OK either way.)
Link to thread (hey, you're already here!):
  https://lists.w3.org/Archives/Public/www-style/2015Feb/0205.html

 (2) "Should min/max size properties be *actively suppressed* when you have to do layout to resolve the flex base size?" (Firefox & IE say "yes, suppress them". The spec used to be explicit about this, but became more vague probably-accidentally, as I noted in the thread.  Chrome disagrees with Firefox/IE & does not suppress the effects of the min/max sizing properties.)
Link to thread:
 https://lists.w3.org/Archives/Public/www-style/2015Feb/0377.html

 (3) "Should min-content size be considered at all, when resolving min-width:auto on an element with an intrinsic aspect ratio?" (Firefox says no; it sounds like IE says yes. The spec used to agree with Firefox, and was then changed -- though it sounded like the rewrite that changed it was just intended to cover more edge cases, not to change behavior like this.)
Link to thread:
 https://lists.w3.org/Archives/Public/www-style/2015Feb/0485.html

> is it possible to have a telecon with the
> editors to resolve these issues?

Works for me! Each of these issues requires some thought & perhaps background/history-reading to grok, so it's possible a telecon will end up with a lot of "hmm, need to think more about that."  But we might be able to knock these out.

> On your above request, I am for resolving percentages
> wherever possible (keeping things the way they are)
> as flex is commonly in a multi-pass state I think the
> intuitiveness for the authors is beneficial.

I'm OK with that, though it makes it pretty easy to hit unfortunate n^2 blowup in deeply-nested-flexbox scenarios, I think. (something like: vertical flex container, which contains a [default] min-height:auto flex item, which contains a percent-height child that is also a vertical flex container, [...continue nesting...]) (Though, it's also possible that there are ways of partially-optimizing this sort of thing away that I haven't discovered/considered.)

~Daniel

Received on Wednesday, 25 February 2015 00:52:03 UTC