[css-gaps-1]: Defer percentages insets to future version of css gaps

[oSamDavis]

In CSS Gaps 1, percentage values used for gap decoration segment endpoint adjustments currently resolve against the crossing gap width at a junction. However, this creates ambiguity and is confusing in layouts where the gap in a direction is not uniform. Specifically, in a multi-line flex layouts where the gap between items can vary due to content distribution/alignment:

This issue proposes deferring percentage values for the inset properties to a future level of CSS Gaps, while we gather more author feedback and converge on an intuitive percentage basis and semantics.

Note: For the most common percentage use case (-50%), issue 13137 resolved to add a keyword. Therefore, the segment joining behavior still remains possible.

cc: @jav099 @kbabbitt

[SebastianZ]

I am fine with deferring, but want to point out that without percentages, transitions between the keyword and <length> values aren't possible. Or are they? Though transitions of gap decorations are presumably an edge case, anyway.

Sebastian

[oSamDavis]

The latest edit of the Spec answers this issue based on the definition of crossing gap width:

Interior segment endpoints
The crossing gap width is the size of the gap junction in the same dimension as the gap in which the segment endpoint lies (the horizontal dimension for row gaps and the vertical dimension for column gaps).
Note: Typically, this will be the used value of the column-gap or row-gap property, whichever applies to the intersecting or abutting gap, plus any additional spacing added due to justify-content or align-content. However, in cases where gaps intersect or abut one another at overlapping intervals, the crossing gap width will be determined by the union of such intervals. See this example for one such case.

Bearing this in mind, we have a clear and consistent way to resolve percentages even in these layouts cases where the gap in a direction is not uniform. So, no need for deference.

[kbabbitt]

The other reason to defer was to reserve percentage basis for something else such as item size and instead assign a keyword for the most common gap-size-relative use case of -50%. But I think the fact that we can have gaps of varying sizes, plus the interpolation issue raised by @SebastianZ, are sufficient to warrant keeping percentages in the spec as they are.