[cssom-view] Clarify about scroll completion

[hiikezoe]

Scroll completion is defined in the CSSOM View spec;

Scroll is completed when the scroll position has no more pending updates or translations and the user has completed their gesture.

But touchpads on MacOS basically send "momentum" events, thus even after the user has lifted their fingers from the touchpad, momentum events are sent to applications.

There's an interesting case we got recently related to this definition, here is the case.

On Chrome, if you try to scroll on the "iframe Scroll Reporter" strongly, you will see a bunch of "type=wheel, button=0, buttons=0 ..." messages even after the scroll position has reached to the bottom edge, then finally you will see a "type=scrollend, target=html" message.

So apparently Chrome considers "momentum events" are parts of a user gesture, but in my naive understanding, "when the user lifted their finger" is corresponding to the "the user has completed their gesture".

[bramus]

But touchpads on MacOS basically send "momentum" events, thus even after the user has lifted their fingers from the touchpad, momentum events are sent to applications.

Isn’t that covered by the ”the scroll position has no more pending updates or translations” part? The momentum events generate pending updates, so as long that happens scroll has not completed.

For clarity, this is how I read the definition:

scrollCompleted = (scroll position has no more pending updates or translations) && (user has completed their gesture)

So when a user swipes down the trackpad, causing a scroll:

1. Upon lifting their fingers from the trackpad, the gesture has completed
2. After the last wheel event comes in, there are no more pending updates/translations to the scrollposition

Only when both 1. and 2. done, scroll is considered to be completed, and scrollend fires.

[hiikezoe]

But touchpads on MacOS basically send "momentum" events, thus even after the user has lifted their fingers from the touchpad, momentum events are sent to applications.

Isn’t that covered by the ”the scroll position has no more pending updates or translations” part? The momentum events generate pending updates, so as long that happens scroll has not completed.

But the momentum events do never change the direction so that once after the scroll position has reached to an edge, there will no pending scroll position updates, I would say.

[bramus]

/ping @DavMila (who implemented scrollend in Blink) for clarification

[DavMila]

Hm, I guess in theory since the update in the momentum event is entirely external to the UA (I think?), if the UA believes momentum events are still incoming, it's reasonable that it behaves as if there are still pending updates.

But my hunch is that in practice these momentum updates cannot suddenly change direction so if you reach the edge, you know there will be no more updates so you can send scrollend - that actually seems more correct to me.

Seems to me that if an author wants, for example, to load more content if the user hits the edge and is relying on scrollend, waiting for the noop momentum events before sending scrollend would be a worse experience.

[bramus]

Seems to me that if an author wants, for example, to load more content if the user hits the edge and is relying on scrollend, waiting for the noop momentum events before sending scrollend would be a worse experience.

Maybe this calls for a new scroll* event?

[DavMila]

Seems to me that if an author wants, for example, to load more content if the user hits the edge and is relying on scrollend, waiting for the noop momentum events before sending scrollend would be a worse experience.

Maybe this calls for a new scroll* event?

IMO it should still be scrollend. Those noop momentum events are not web-observable I believe.

[whimboo]

When testing with the referenced testcase, I noticed that doing a quick flick (near the end of scrolling within the iframe) produces a series of wheel events where deltaY increases and then decreases, while pageY remains constant. I assume these are momentum scroll events, which suggests they're exposed to the web?

In Firefox, it looks like we fire the scrollend event immediately after scrolling truly stops, ignoring those momentum events — unlike Chrome, which seems to include them before sending out scrollend. Please correct me if I’m misunderstanding.

In Safari I do not see scrollend events at all, so I assume it's not implemented yet.