[css-gaps-1]: Inheritance of gap decorations within subgrid

[oSamDavis]

According to the CSS Grid spec, when a subgrid does not define its own gap property, it inherits the value from its parent grid. Considering this, should gap decorations also be inherited by default in a subgrid context?

[kbabbitt]

One note, the gap sizes are not "inherited" in the CSS sense. Rather, the normal keyword is defined, as I understand it, such that the used value will be the same in the subgrid as in the parent grid:

A value of normal indicates that the subgrid has the same size gutters as its parent grid

For gap decorations, I don't think we can just use the same value directly; it seems unlikely to be what the author wants if they've specified varying decorations in the parent grid. For example, suppose the author specifies column-rule-width: thick repeat(auto, thin) on the parent grid to get a thick line in the first column gap only. They would also get a thick line in the first column gap of the subgrid, which is not going to line up with the thin lines in the rest of the parent grid. (Unless of course the subgrid happens to occupy the first two columns of the parent grid.)

Instead, we'd need to do something where we line up the decorations similar to how Grid 2 lines up the gaps themselves. That feels kind of magical to me, but if it happens for the gaps, I can see authors expecting it to happen for the decorations as well.

[oSamDavis]

Thank you for your response and clarification on the "inheritance" behavior. You made a great point that we would not want to blindly inherit gap decorations from the parent grid as it might not align with the author's intent. However, there might be a case for selective or "smart" inheritance, where decorations are inherited only for the specific gaps corresponding to tracks a subgrid spans.

To examplify, I have:

In the above, the subgrid inherits the decorations of the gaps it spans. I'm not certain this behavior is desirable from a web author's perspective, but it's worth noting the discussion in 2285, which motivated the normal keyword for gap inheritance primarily to reduce syntax repetition in subgrids. That rationale could arguably extend to gap decorations as well.

[tabatkins]

I agree that, in general, inheriting gap decorations into subgrids isn't desirable. A subgrid is usually an independent object that is just using the parent grid to ensure that it aligns with its sibling subgrids.

I could see a potential use-case for explicitly opting in to sharing gap decorations, but I'd like to see them in use first to see if people end up complaining about the lack.

[kbabbitt]

I agree with Tab. My suggestion would be not to do anything for this scenario in Level 1, but rather capture it as a future idea in the explainer and look at in the future if there's author demand.

@oSamDavis wdyt?

[oSamDavis]

I agree as well. This can be done in the future if the demand arises.

[kbabbitt]

@oSamDavis I added a section to the explainer to capture this idea: MicrosoftEdge/MSEdgeExplainers#1117

I wasn't able to add you as a reviewer to the PR, but does that look good to you?

[oSamDavis]

Yes, that makes sense and looks good to me. Thank you!