[css-grid-3] masonry-auto-flow has suprising default

[bfgeek]

https://drafts.csswg.org/css-grid-3/#masonry-auto-flow

masonry-auto-flow has the pack default which will explicitly pull items out of order via definite-first (I believe). This is suprising! (Also definite-first isn't really defined).

At the very least this should be ordered by default, but likely without good justification the definite-first concept should be removed.

Additionally use-cases for masonry-auto-flow: next would be good, I personally haven't seen web developers ask for this behaviour - likely next could be dropped as well.

[Loirooriol]

Additionally use-cases for masonry-auto-flow: next would be good

Having a stable layout that isn't completely rearranged when some item changes size. Seems useful to me.

[bfgeek]

Having a stable layout that isn't completely rearranged when some item changes size.

Its bad for accessibility however - as it pulls things into a weird ordering. If things are mostly the same size grid should suffice, or if its about small reordering errors, then the reorder-threshold property should suffice.

[Loirooriol]

It doesn't seem a weird ordering to me. It obviously has the gotcha that it can fill the columns at very different paces, but this doesn't make it weird, and it isn't a big deal if e.g. you have gazillions of contents available to lazy load whenever the user reaches the end of the shortest column.

In fact, even though pack has some clear advantages, I would say that it produces a weirder ordering. And precisely reorder-threshold tries to mitigate that. But reorder-threshold doesn't bring stability: there is still a threshold that will produce one result when the size is of item is below it, and a completely different one when it's above it.

[tabatkins]

Re: next vs pack - yeah, Oriol's argument makes sense. If you have some <details> in your masonry, and you don't want the layout to jiggle when you open one, next will keep things nice and stable. Your layout will be worse, but it'll be stable, and I think that's a tradeoff that's reasonable for authors to make sometimes.

(On the other hand, as noted elsewhere, using masonry-threshold: calc(infinity) is identical to pack, so we might be able to get away with just using that, perhaps with a note to authors about it. But I'm not opposed to keeping the keyword if it's a more straightforward/learnable way to express this, and then having a note that it makes masonry-threshold powerless.)

[fantasai]

We updated the default and dropped masonry-auto-flow per https://www.w3.org/mid/CADhPm3s2EftVNQpRWmw3gyCEFB_JYdW5F=Gku2ccd4ADBO60EA@mail.gmail.com