[css-values-4] `round(line-width, ...)` specification is confusing

[romainmenke]

see:

• f63d90e
• https://drafts.csswg.org/css-values-4/#round-func

Otherwise, identical to nearest, except that if A is positive (non-zero) and one of lower B or upper B is zero, choose the non-zero one, then snap as a border width.

Can be read as:

• line-width is almost identical to nearest
• has an exception when lower/upper B is zero and ...
• always snaps to border width

This violates the assert in https://drafts.csswg.org/css-values-4/#snap-a-length-as-a-border-width

Assert: len is non-negative.

or:

• line-width is almost identical to nearest
• has an exception when lower/upper B is zero and ... only when in this exception -> snap to border width

When upper and lower B are both non-zero and A is positive (non-zero) there still isn't any border width snapping. This seems weird to me.

---

If <rounding-strategy> is line-width, B may also be omitted and has special behavior (defined above) essentially defaulting it to one device pixel. In all other cases, omitting B is invalid.

Can be read as:

• line-width allows B to be omitted
• always has special behavior -> snapping to border width

or:

• line-width allows B to be omitted
• only when B is omitted the value is snapped to border width

---

@tabatkins I assume the intention is to always have the border width snapping behavior?
It might help to split into a bit more sentences :)