[css-values] Should round(strategy, A, 0) be NaN or A?

[Loirooriol]

Currently https://drafts.csswg.org/css-values-4/#round-infinities says

In round(A, B), if B is 0, the result is NaN.

But according to @tabatkins "my aesthetic sense is telling me I'm not allowed to say it's NaN when there are clear limit values."

In this case the limit is clear: lim_{x→0} round(strategy, A, x) = A.
So maybe round(strategy, A, 0) should be A?

On the other hand, being A would seem to imply that A is a multiple of 0, which technically it's false.
But A / ∞ = 0, so in a sense we do have A ≈ 0 * ∞, dunno.

[Crissov]

round(A, zero) should intuitively equal round(A) and thus A indeed.

[Loirooriol]

@Crissov Note the spec doesn't allow round(A), and if it did I would expect it to be round(A, 1) (i.e. round to integer), not round(A, 0).

I don't have a strong opinion about round(A, 0) being A or NaN, but the choice doesn't seem obvious, so I filed this so that people can give some thought to it.

[tabatkins]

Yeah, in this case the limit is different from the behavior at the limit, I think. The invariant that the result is a multiple of the step is broken.

But A / ∞ = 0, so in a sense we do have A ≈ 0 * ∞, dunno.

Note that 0 * infinity is NaN. ^_^

[Loirooriol]

Sure, that's why I said "in a sense", i.e. it's one of the possible solutions of the indetermination, but there are others.

[tabatkins]

I'm going to go ahead and close this no-change. I can see the limit argument for making round(A, 0) yield A, but the invariant argument for making it NaN (the result should be a multiple of the second argument, and nothing non-zero is a multiple of 0) still seems the most reasonable to me.

@Loirooriol, does this work for you?

[Loirooriol]

Yes