Event path trimming should only happen for mouseover and mouseout

[rniwa]

It's very strange for the event path trimming algorithm to apply trusted events with relatedTarget.

Why don't we just do this for all mouseover and mouseout events and nothing else. Alternatively, we can add a new constructor option to MouseEvent's init dictionary like stopsAtLowestCommonShadowRoot to explicitly set this behavior so that authors can adjust it. If we did the latter, we'd just make UAs create MouseEvents with that flag set to true when firing trusted events.

@travisleithead @annevk: any opinions?

[rniwa]

If we (in some off chance) were to keep the currently spec'ed behavior, then the following text belongs to an informative note, not a normative text since this trimming behavior applies to all trusted events.

to avoid the appearance of spurious mouseover and mouse out events firing from the same node

[annevk]

Doing this flag-based does seem more sane. Since it would actually explain what is going on. (Where did I hear that before?)

[rniwa]

Yeah, I think I like the flag-based approach best. cc @hober @othermaciej

[hayatoito]

This is the original bug for the current spec. https://www.w3.org/Bugs/Public/show_bug.cgi?id=20017

The flag-based approach is okay to me.
We need a better name than stopsAtLowestCommonShadowRoot. Any idea?

[rniwa]

An alternative to specify a boolean would be to specify a node at which the propagation stops: e.g. highestTarget. This way, mouseover and mouseout can supply the common ancestor as highestTarget instead. This feature will be slightly more powerful than what's needed by this use case. Although I'd imagine it could become handy in some cases, I'm not sure if it's worth the complexity.

[hayatoito]

I'm not sure how we can resolve this issue. Does someone have a preference?

[rniwa]

My preference is to add a flag instead of treating trusted events differently since the rationale for trimming the event path is semantically nothing to do with whether event should be trusted or not.

[annevk]

untilCommonShadowRoot?

[hayatoito]

I think we agreed on using the flag. Let me update the spec so that it uses the flag in relevant sections.

Note that the flag name is not finalized. It could change if we find a better name.