- From: Tab Atkins Jr. <jackalmage@gmail.com>
- Date: Mon, 3 Feb 2014 15:34:02 -0800
- To: Julien Chaffraix <jchaffraix@google.com>
- Cc: www-style list <www-style@w3.org>
On Wed, Jan 29, 2014 at 5:41 PM, Julien Chaffraix <jchaffraix@google.com> wrote: > Hi, > > * 'flex-end' resolves to 'start' on non-flex items > > This is very confusing and I think it would make more sense to have it > resolve to 'end' to be consistent with the author's cue. The reasoning for this is that if you're typing 'flex-end' and you're not in a flexbox, we have no idea what you actually intend. Our usual approach is to error-correct into something close to your intent if we can reasonably infer that, and just reset to the initial value otherwise. That said, I don't actually have a strong opinion here. If you say 'flex-end' accidentally, it might be *reasonable* to assume that you meant to say 'end'. I'm okay with making this change. Somewhat unrelated, I note that 'flex-start/end' are defined to be "equivalent to" start in a non-flex context. We should probably just compute them to start instead; doing so relies on information that is generally okay to deal with in computed values (just the 'display' of the element's parent). > * Currently the specification is silent on what happens when > 'self-start' and 'self-end' are set on an orthogonal writing mode. > > I have thought of 2 ways to think about this (there is probably others): > A) As the axes from the containing block / child are orthogonal, it is > invalid and we would default to 'start' / 'end' (based on the original > property). > B) We use the child's coordinate system to resolve start / end into a > physical direction and use it for the resolution. > > >From my perspective, A) makes more sense as B) would involve looking > at the opposite axis (e.g. 'justify-self' would end up working on the > child's block-axis). It's not silent; the behavior is well-defined, it's just stupid. A grid item using "align-content: self-start;" and an orthogonal writing mode would just align its inline-start edge with the corresponding block-* edge of the grid-area. I'm okay with changing to your suggested behavior A, and just computing it to start/end instead. ~TJ
Received on Monday, 3 February 2014 23:34:50 UTC