[css-shapes-2] Disallow interpolation between `path()` and `shape()`

[smfr]

Basic Shapes 2 says that path() and shape() should interpolate.

I don't think that's very extensible going forward, where we're likely to add new shape commands that don't correspond to types that can be interpolated with SVG-type path commands, and it doesn't seem particularly useful for authors.

I suggest we disallow interpolation between path() and shape().

cc @noamr

[noamr]

Yea I think it's not that necessary to have path/shape interpolation.

[tabatkins]

I don't understand this objection. Having different value spaces between the two types isn't a problem, so long as one is a superset of the other, and afaik that's precisely what we intend to preserve - every single behavior you can express with path() should be expressible with shape(), and shape() might grow additional commands that we don't sync back with SVG.

The intermediate value would just be shape() - you'd "upgrade" the path() into an equivalent shape() and then interpolate using shape() rules. This sort of behavior has several existing examples in CSS.

[smfr]

It just seems like unnecessary implementation complexity. If an author is using shape() in one state, it's likely that they've converted their path()s to shape()s already.

[ydaniv]

I agree with @tabatkins, you should be able to upgrade path() and make this seamless to authors. And not unnecessary if you consider the entire ecosystem is still filled with path()'s coming from SVG/CSS and tools that generate them. IMO being able to work seamlessly with both will be very valuable.

[noamr]

Personally I am OK either way.
@smfr I don't see a problem with future extensibility; path() can always be read like a shape() with px values, and then interpolation can continue from there according to the shape() rules.

[smfr]

OK, I'll close.