It might even be clearer with the following two slightly simpler examples:

<div style="margin-left: 10px; margin-right: 20px; direction: rtl">
  <div style="margin-left: inherit /* compat requires 10px */; direction: ltr"></div>
</div>

<div style="margin-left: 10px; margin-right: 20px; direction: rtl">
  <div style="margin-inline-start: inherit /* 10px or 20px? */; direction: ltr"></div>
</div>

In other words, does inherit inherit the computed physical value from the parent (always), or does it inherit from whichever of physical or logical was specified?

I would hope, at least, that those two examples produce the same results (respectively) as these two:

<div style="margin-inline-end: 10px; margin-inline-start: 20px; direction: rtl">
  <div style="margin-left: inherit; direction: ltr"></div>
</div>

<div style="margin-inline-end: 10px; margin-inline-start: 20px; direction: rtl">
  <div style="margin-inline-start: inherit; direction: ltr"></div>
</div>

since the outer div always has the same computed values for all properties (the difference in that in these last two, it got those values through logical properties rather than physical ones).

Just wanted to note here that the possible solutions to #3030 depend on what happens here.

I would hope, at least, that those two examples produce the same results (respectively) as these two.

Yes, they would have to: inheritance uses the parent’s computed value, and pairs of corresponding logical/physical properties are required to have the same computed value.

What about margin-inline-start, though? Does it inherit 10px or 20px?

Chromium, WebKit and Firefox are interoperable, it inherits 10px. So if margin-inline-start resolves to margin-left, it inherits the margin-left of the parent, even if that's the margin-inline-end of the parent.

@Loirooriol Thanks! We should probably check with i18n that it's acceptable behavior, but as long as it is, speccing interop sounds reasonable to me.

fwiw, last Friday i had created about 100 tests for css-logical, including the various directional permutations. It's geek week now, so i may upload the tests with results this week, or it may have to wait until next week. I might be able to add a test for this scenario, too.

So the main body of tests and results is at https://w3c.github.io/i18n-tests/results/css-logical

I didn't yet integrate the tests for inheritance into that page, because i'm not sure about the result. You can find 2 tests at

• https://w3c.github.io/i18n-tests/css-logical/inheritance/inherit-logical-logical.html
• https://w3c.github.io/i18n-tests/css-logical/inheritance/inherit-physical-logical.html

The results are indeed interoperable across Firefox, Chrome, and Safari browsers.

However, i wrote the assertions following the logic that seemed sensible to me: that is that the margin-inline-start for a LTR block (ie. on the left) would inherit the value for the margin on the right (ie. at the start) of a parent RTL block. Under that criterion, all the tests fail.

It seems somewhat illogical, as it were, that what is being inherited by a logical value is not the start (ie. logical) value of the parent. It would mean that a content author has to do some mental gymnastics to decide whether or not to use inherit explicitly.

@r12a would you be available for any of our remaining meetings this week to discuss this issue?

https://wiki.csswg.org/planning/tpac-2020#schedule

So these are the two operations under considerations:

• Blue Line: Child’s margin-inline-start maps to margin-right which then inherits from parent’s margin-right (which is the parent’s margin-inline-end). This is what browsers currently do.
• Red Line: Child’s margin-inline-start (which is the child’s margin-right) inherits from parent’s margin-inline-start (which is matched with parent’s margin-left). This is what @r12a expects.

Keep in mind that properties are first cascaded (all together, logical and physical mixed and mapped by writing mode, latest declaration of a conflicting set winning), and then computed (which causes each matching pair to share a single computed value).

Note that if logical properties inherit from logical properties, we need to define what happens with all: inherit, since the order will matter. For backwards compatibility it should either not include logical properties, or it include them before physical ones.

I think it would be simpler to just keep what's currently implemented (inherit from the resolved physical properties).

@Loirooriol We resolved yesterday that shorthand properties only apply to one type of physical-or-logical mapping.

During parsing of a css declaration block, shorthands (like margin and all) expands only to the set of longhand (grouped by mapping logic: so either logical or physical) required to describe their value, and by default resort to use physical properties if both sets are possible. Note that they can only expand if they do not contain var() substitutions.

The CSS Working Group just discussed css logical properties ordering, and agreed to the following:

• RESOLVED: inheritance of logical properties inherts the corresponding logical proeprty on the parent

@fantasai I don't really get this. Which "omitted values"? "otherwise" which condition?

@Loirooriol If there's a shorthand that can set both physical and logical values, e.g. suppose we extended margin in such a fashion, then whichever longhands are omitted (not set explicitly) should be set first so that the explicit values reset whatever is necessary. So if someone specifies margin: relative 1em then we set the logical longhands last so they can win against the physical ones. If we always set physical last, then setting logical values in such a shorthand would be impossible.

@fantasai But according to #3030,

During parsing of a css declaration block, shorthands (like margin and all) expands only to the set of longhand (grouped by mapping logic: so either logical or physical)

while you seem to mean that both sets of longhand are set? What is supposed to happen if I do

document.body.style.cssText = "margin: 0";
document.body.style.length; // 4 or 8 ??

My understanding is 4, the logical properties only. But this is understable since such property doesn't exist yet...

Well, margin already exists. When not including the non-stable logical keyword, should it only expand to the physical properties only, or should it also reset the logical ones beforehand? From @fantasai's explanation the spec seems to want the latter, but I'm not sure if this is web compatible.