[css-position-3] Define containing block for stickypos internal table elements

[fantasai]

Internal table elements don't really have a "containing block" defined per se, but one is needed to intersect with the viewport for sticky positioning. So we need to define that, but also what should it be? The table grid box? the table row box? something else?

[tabatkins]

I think the answer should be that the containing block for rows and cells should be their row group, and for row groups it shoudl be the table grid box.

This way you can get sticky headers with thead { position: sticky; top: 0; }, but also get individual sticky rows that can get "carried away" by judicious use of tbodys, just like sticky headings + section.

[atotic]

It's the "sticky failures" day at TablesNG manor.
Here is a test case: https://wptest.center/#/1vb898 (my first time using wptest)
It applies sticky constraints table parts, scrolls, and observes part position.

I am testing for the most common behavior in current browsers which is:
"Table wrapper fragment is the sticky container for all table parts"

Results:

// FF 82: pass
// Chrome legacy: TD pass, TR/TBODY fail, sticky treated as static
// Edge 17:TD pass, TR/TBODY fail, sticky treated as static
// Safari: TD/TR fail creatively, TBODY pass.

For TD, sticky container being Table wrapper matches widest current behavior.
If we adopt this for TD, makes sense to do it for TR too.

[jonjohnjohnson]

I think the answer should be that the containing block for rows and cells should be their row group, and for row groups it should be the table grid box.

@atotic Are we going to run into compat issues by delaying a possible resolution on @tabatkins suggestion?

[atotic]

I do not think there will be compat issues with tab's proposal.

Chrome has shipped it, and developers are happy.

[tabatkins]

@atotic Are you saying that Chrome has shipped the "cells/rows stick to their row-container" behavior I suggested above, or just that Chrome has shipped the "all internal table elements stick to the table wrapper" behavior? When I play around with your test from above, I'm seeing cells only stick to the table as a whole.

[atotic]

What I implemented was table wrapper is a container for row-groups/rows/cells.
My comment "I do not think there will be compat issues with tab's proposal." is incorrect.
I got confused, and forgot that that tab's proposal was "row-group is a container for rows/cells".

[jonjohnjohnson]

@atotic

My comment "I do not think there will be compat issues with tab's proposal." is incorrect.

Do you still think there will be compat issues? If so, should this be discussed/resolved sooner than later to lessen the impending issues? What are the next steps in your opinion?

https://bugs.chromium.org/p/chromium/issues/detail?id=702927