[web-animations-1] Keyframe computed value computation is confusing

In https://drafts.csswg.org/web-animations-1/#calculating-computed-keyframes the spec tries to redefine how the cascade works, for example, saying that longhands override shorthands. This seem to imply that, e.g. in

@keyframes cascade-test {
    from {
      border-style: double;
      border: solid;
    }
    to {
      border-style: dashed;
      border: solid;
    }
  }

we would get an animation of a dashed/double border, instead of a solid border (as a true cascade would imply).

It's only reasonable to apply these rules to an API that takes a set of property declarations without ordering them.

IMHO, WA1 should only define a rule for ordering an unordered set of property values, and say that it only triggers if there is no inherent ordering in the API using it. And other than defining an order, it shouldn't try to interpret the properties, instead leaving that to CSS to define. (For example, WA1 gets the input to mapping between physical and logical properties wrong.)

Start with the “Calculating computed keyframes” section linked in the issue and trace how the provided @keyframes example is handled. Compare the current cascade interpretation with the proposed treatment of unordered property values, including physical and logical properties. Done means the Web Animations Level 1 wording clearly defines ordering without incorrectly interpreting CSS properties.