[css-align] don't introduce baseline alignment to table columns

[dbaron]

The section on Baseline Content Alignment and the definition of alignment context introduce a concept of baseline alignment to table columns, which can happen when the writing mode of the cells is orthogonal to that of the table. (The parts in-between are defined (by the Applies to line of writing-mode to have the same writing mode as the table.)

I think this is a very complicated feature, and should not be introduced.

Table layout is a rather complicated algorithm that is not particularly symmetric. Column size calculation happens first, followed by layout and row size calculation. Introducing baseline alignment to the column phase would be a complicated change, and it's not clear to me how it would work. It is more complex than the union of:

• introducing character alignment
• doing intrinsic sizing of orthogonal table cells
  since it has most of the issues that both of those problems have, and then I think some of its own.

I don't think the complexity is justified without substantial evidence of demand.

[dbaron]

Also, see the details on character alignment, with issues in text level 4.

[FremyCompany]

I totally agree here.

I am not even sure such a thing would be possible, because laying out the first line in the column direction for table-cells using an orthogonal writing mode would require to know the height of the rows (because this will change how many characters can fit in the first line, and if characters are sorted in constantly increasing font-size, each new char will push the baseline further down). Since the height of the rows can only be computed once the width of the columns has been decided, so we have a circularity here. (a similar argument might actually apply to grid as well, but I am not sure).

Besides, the current situation with orthogonal writing modes in tables is absolutely terrible, we don't
even have a base on top of which to build. These two test cases should convince you:

https://wptest.center/#/m2f950 :

screenshots

https://wptest.center/#/ie2h39 :

screenshots

---

Also, why does justify-self still allow baseline alignment? I thought baselines were only generated for the inline direction, so you could not possibly justify yourself based on a baseline, can you?

[tabatkins]

Also, why does justify-self still allow baseline alignment? I thought baselines were only generated for the inline direction, so you could not possibly justify yourself based on a baseline, can you?

align vs justify is based on the container's writing mode - if your container is horizontal but you're vertical, you can absolutely justify yourself to a baseline.

[FremyCompany]

Also, why does justify-self still allow baseline alignment? I thought baselines were only generated for the inline direction, so you could not possibly justify yourself based on a baseline, can you?

align vs justify is based on the container's writing mode - if your container is horizontal but you're vertical, you can absolutely justify yourself to a baseline.

Oh, I see, that makes sense.

[css-meeting-bot]

The CSS Working Group just discussed don't introduce baseline alignment to table columns, and agreed to the following resolutions:

• RESOLVED: The baseline alignment in block directon of tables is not defined. Also add a note prompting impl to bring back ideas if they have any

The full IRC log of that discussion

<dael> Topic: don't introduce baseline alignment to table columns
<TabAtkins> (Rather, not specified values, just *not* the result of the overconstrained-equation.)
<dael> Github: https://github.com//issues/1422
<dael> Rossen_: I think there was less compat on this.
<tantek> Agenda+ https://github.com//issues/1002 (if we have time for it)
<dael> fremy: dbaron mentioned in alignment spec there's a line that says cell in a table sharing a column are in same alignment group if there's in the alternate writing mode of the table. Basically dbaron said it seemed difficult to impla nd I agree so he proposed to remove this from align spec
<dael> fremy: It seems difficult to me to get the result, what dbaron said seemed right to me. At this point there's no browser that can handle vertical left right cells properly anyway so I think we can go ahead and baseline align.
<dael> fremy: I agree with dbaron
<dael> dbaron: I think one other analogy is the complexity is similar to impl character alignment in tables. It's in substantially higher demand and also not impl. Seems like the complexity is similar, but it would complicate thigns to do both. I'd rather character alignment happens first
<tantek> perks up and agrees with dbaron re: the complexity of character alignment (and non-implementation thereof)
<dauwhe> +1 on priorities :)
<dael> fantasai: I understand the priorities here. What if the spec said MAY so if an impl wants to do this they can.
<dael> Rossen_: One thing to point out is we had a similar discussion when working on grid for alignment of grid items and there we had a strong vote to not baseline align on the minor/block direction.
<dael> Rossen_: Since baseline align is an inline concept trying to align to those is fairly tricky. When you throw this into the mix of all the dependency I'd hav to agree with dbaron that making this work and work interop will be a tall order.
<fantasai> s/dependency/sizing dependency/
<dael> Rossen_: I would side on not reqiring this as an impl.
<dael> myles: Everything you said is correct Rossen_
<dael> Rossen_: fantasai you seem to have been the one wanting to try and make this work. How strongly do you feel?
<dael> fantasai: If the main concern is impl complexity I think it makes sense not to require. If there was an impl that decided it was imporant for their use case like if they have a lot of mixed writing modes then that impl should be allowed to do that without violating spec. I want to make this optional and we can make it clear we're not expecting impl but they're allowed to
<dael> dbaron: You're saying make it optional but don't spec how it works since how it works changes the table algo pretty deeply.
<dael> Rossen_: I was also going to ask about making it explicitly non-defined rather than not allowing it would let the impl come back to the table. Could we make a way for this to work in all our frameworks. If we make it strongrer it will give those impl that are violating to come back and say this is important and here is how it should work.
<dael> Rossen_: fantasai would you agree or prefer it to be in spec and a may?
<dael> fantasai: I suppose that's fine.
<dael> Rossen_: Other opinions?
<dael> Rossen_: Proposal: The baseline alignment in block directon of tables is not defined.
<fantasai> "This may be added in a future level if there is sufficient demand and implementer interest" ?
<dael> RESOLVED: The baseline alignment in block directon of tables is not defined. Also add a note prompting impl to bring back ideas if they have any