[css-position-3] Circular definition of fixed positioning containing block

[vwkd]

(For simplicity, I'll assume continuous media below.)

The definition of the fixed positioning containing block in css-position-3 reads

If no ancestor establishes [a fixed positioning containing block], the fixed positioning containing block is [...] the viewport.

This definition is circular. You can't define "fixed positioning containing block" assuming its very existence in the preceeding condition "If no ancestor establishes [an fixed positioning containing block]".

Also this contradicts the comment only a few lines above, which defined it directly as the viewport.

[...] a fixed positioning containing block (the viewport [...]).

Furthermore, ignoring the contradicting comment above, the term "fixed positioning containing block" is left completely undefined.

Currently, it seems a unnecessary complication for the definition of the term "containing block of fixed positioned elements". Instead of defining A as C ("containing block of fixed positioned elements" as "viewport"), the spec defines A as (B or C) ("containing block of fixed positioned elements" as "fixed positioning containing block" or "viewport") and leave B entirely undefined. Furthermore, the choice of the term B - "fixed positioning containing block" - contains the very words for whose definition it's used - "containing block" -, which is to say the least bad naming. It's as if you define the word "car" as "an object with a car door" and then need to define "car door" without using the word "car".

It should be noted, that an identical argument can be made for the term "absolute positioning containing block".

[Loirooriol]

I guess it should just be

The containing block is established by the nearest ancestor box that establishes an fixed positioning containing block [...]
If no ancestor establishes one, the fixed positioning containing block is: [...]

[vwkd]

@Loirooriol I agree, this is one way to a fix.

This implies that the "fixed positioning containing block" is not the viewport, meaning the comment in the parentheses would need to be removed as well.

[...] a fixed positioning containing block (the viewport [...]).

But it still leaves the term "fixed positioning containing block" completely undefined. And it doesn't fix that using a more special term to define the more general term is wrong, just like "car" can't be defined using the term "car door" since it relies on the very term "car" that it is trying to define.

[tabatkins]

You're right that the parenthetical was incorrect; I've amended it to say "usually".

However, the definition of FPCB is not circular. The reference in question is to "establishes a fixed positioning containing block", which is a term of art that some properties invoke; this section defines what the FPCB is when this happens (a rectangle determined identically to the abspos containing block on the element). If no element "establishes a FPCB", then the definition defines how to calculate the FPCB from the viewport/page instead. That's a well-founded definition.

Is this acceptable?