Element parameter as the presentation area might make it hard to implement

[NavidZ]

In the requestPresenter API we have an element as the presentation area meaning that we will never draw beyond the boundary of that element with this API.
I wanted to bring this up that an element on the web could be randomly shaped with all sort of clip paths.
We should be able to support that with extra plumbing within Chrome but I wonder if the native APIs such as the one proposed on Windows could handle such a thing.

[dlibby-]

I was writing up a response how in practice this shouldn't matter, since clip-path should affect hit testing and we would not produce a trail if the website does not call setLastRenderedPoint during a frame where a PointerEvent was dispatched.

But I failed to realize the implications of implicit pointer capture, where move events are streamed to the listener that received pointerdown.

I think conceptually having it be an axis-aligned bounding rectangle could be a reasonable limitation (and indeed the proposed Windows API only handles rectangles), but this does need more thought.