In absolute positioning in general, if an author sets one inset and leaves the other one 'auto', we attach the absolutely positioned box to that edge (by magically resolving the auto inset to an appropriate length to make the containing block snugly fit the box into that position). But with position-area, we resolve all 'auto' insets to zero, which means that we lose this behavior.

There are two downsides to this:

• It's likely unexpected, since this works elsewhere. Usually the default alignment is what's wanted, but for center tracks, there's a little more room for preferences.
• The alignment properties are fundamentally logical, but position-area has options for fully logical, fully physical, and logical-physical placement combinations, so controlling the position within an area with alignment properties is not always ideal.

We could consider altering the alignment behavior for position-area so that in the cases where there is only one auto inset, we still resolve that to a used value of zero (leaving us with a usefully-sized containing block, if the author changes the alignment), but also resolve normal alignment to attach to the non-auto edge instead of to its usual area-based default.

What do people think? CC @kizu @mirisuzanne