restored definition of automatic block sizing pulls in issues related to intrinsic sizes in the block dimension

[dbaron]

I have one concern about the primary paragraph that was just restored in #4220.

The text of this paragraph seems to unnecessarily redefine the automatic size in the block dimension in a complicated way that pulls in all of the issues in #2890.

It seems to me that the point of the paragraph is to produce a different automatic size in the inline dimension (which is the block dimension of the parent), which is the dimension in which the min-content and max-content concepts are clearly defined. I think it's quite reasonable in that dimension.

In the block dimension, I think the intent of the paragraph is to produce the same rules as the default behavior: the automatic size is the result of layout (since I think the min-content and max-content sizes in the block dimension are supposed to be the same, which makes the fit-content expression given equal to both of them). But it does that by referring to concepts that aren't clearly defined (see #2890) and which might be interpreted in other ways.

I would prefer if this paragraph instead just redefined the automatic size in the inline dimension and left the automatic size in the block dimension alone.

(That said, I'm thinking about this primarily for a display: block orthogonal flow, and haven't really thought through what the automatic sizes are for flex or grid.)

[fantasai]

“and left the automatic size in the block dimension alone” means what, exactly?

[dbaron]

deferring to the existing definition of automatic block size that already exists for its display type

[dbaron]

IRC discussion

[fantasai]

Verified via IRC