[css-align][css-tables][css-flex] Synthetized baselines seems like a breaking change

[FremyCompany]

The new specs states that baseline alignment shouldn't happen to synthetized baselines when alignment happens in a container with a top and bottom instead of just a text line (something along those lines).

This is a breaking change for flexbox and css-tables, and I think we should talk about this now.

---

Also, about synthetized baselines, I understand tables align to baselines even things that have a synthetized baselines, and I don’t think we wanna change that. Maybe we need to clarify that somewhere?

Yes, we need to do that.

Where do you think we should write that statement? In the table spec, or as part of the changed text you introduced?

I am also worried there might be compat issues for flexbox that would prevent us from doing the change there. I think that's a question we should bring up to the Working Group for consideration.

[FremyCompany]

https://jsfiddle.net/qjmjnz5o/2/

[FremyCompany]

The distinction was mainly intended to ease issues related to block-aligned axis. Since that feature was dropped, we can now safely close this issue as well, with the resolution that we will continue to synthetize baselines on everything in the inline axis as we did before.

We resolved so in today's working group telcon.

[fantasai]

Reopening because this isn't entirely resolved. The current situation is as follows:

• Flexbox is underspecified wrt items without baselines (in particular, flex items with display: block don't have a defined baseline).
• Flexbox requires items that have a writing mode orthogonal the baseline-alignment axis to be flex-start aligned.
• The baseline of a flex container itself is synthesize from its first item’s content box (or its own content box, if it has no items).

https://drafts.csswg.org/css-flexbox/#flex-baselines

• Firefox aligns items without baselines and those with orthogonal writing modes at their bottom margin edge.
• Chrome aligns items without baseline at the bottom border edge and uses start-alignment for orthogonal writing modes.
• Edge has the same behavior as Chrome [edit by Rossen]

http://software.hixie.ch/utilities/js/live-dom-viewer/?saved=4416

There's a few concerns here:

1. Probably, boxes in orthogonal flows and boxes without a natural baseline should be treated the same.
2. It's uncertain whether authors want such items to have a “baseline” alignment, or whether merely having them fall back to start alignment would be sufficient.
3. The inline-layout practice of using the margin edge is useless when the margin is positive because it does not align anything visible to the baseline. The author could use negative margins to control the alignment point, but that risks overlapping content.
4. In grid layout, we have items that can span multiple rows; in this case baseline-aligning the bottom edge of the item could be awkward.

Further notes:

• The alignment point of a box without a baseline set can be controlled with the alignment-baseline property; it will use the top/center/bottom as appropriate. Which edge is used (margin/padding/border/content) is not controllable, however. A future property may allow more precise control of how baselines are synthesized, though.
• Inline layout synthesizes a baseline from the margin edge of an item. Note it performs baseline self-alignment (shifting the box itself).
• Table layout synthesizes a baseline from the content edge of a cell. note it performs baseline content-alignment (adding extra padding to shift the content within the box).

Proposals:

• A. Make boxes without a natural baseline and boxes establishing an orthogonal flow synthesize a baseline.
  • A.1 Base this synthesized baseline on the margin box edges
  • A.2 Base this synthesized baseline on some other box edge
• B. Make boxes without a natural baseline and boxes establishing an orthogonal flow use start alignment instead of trying to synthesize a baseline.
• C. Something else.

Questions to consider:

• Do authors need or want synthesized-baseline alignment? If not, falling back to start is simpler.
• Are there considerations wrt implementation complexity?

[FremyCompany]

I thought we resolved on A already during last telcon?

For baseline alignment the conclusion was we keep synthesizing baselines for everything

Except if we really want rediscuss that, that leaves A1 vs A2. I guess for the sake of consistency, we should probably just have the behavior of flex be the same as a table-row (and therefore use margin box?), though I have no strong opinion on that.

[astearns]

Not sure it's ready for the meeting again - please add the tag back if it is

[fantasai]

Table rows don't use the margin box, they use the content box of the cells.

[fantasai]

And no, we didn't resolve on this. That was a comment from me about our conclusion, but it was not a WG resolution. Comments from some random WG member in a meeting, even if minuted, do not constitute a WG resolution. WG resolutions are things that are specifically and explicitly called out in a RESOLVED line.

[FremyCompany]

Just adding this test case for tables, since I wrote it anyway:
https://jsfiddle.net/8c88yb7q/1/

[fantasai]

CSSWG resolved on A.2 - synthesize from border box.
https://lists.w3.org/Archives/Public/www-style/2016Sep/0000.html