`movementX` and `movementY` for `pointermove` (and `pointerrawupdate`) should be the difference from the previous event for the same `pointerId`

[masayuki-nakano]

It's defined as having non-zero values for pointermove and I suggested that they should be available with pointerrawupdate too.

• Unclear if pointermove should set movementX/Y #100
• Define pointermove and pointerrawupdate may have delta from previous corresponding pointer event as movementX and movementY pointerevents#535

However, the Pointer Lock 2.0 spec does not mention about pointerId. I think that the values should be difference from the corresponding event whose pointerId is the same as the event.

E.g., when there are 2 pointers,

1. pointermove for pointerId: 1
2. pointermove for pointerId: 2 (should be zero)
3. pointerrawupdate for pointerId: 1 (should be diff from the first pointermove)
4. pointerrawupdate for pointerId: 1 (should be diff from the previous pointerrawupdate)
5. pointermove for pointerId: 1 (should be diff from the first pointermove)
6. pointerrawupdate for pointerId: 2 (should be diff from the second pointermove)
7. pointerrawupdate for pointerId: 2 (should be diff from the previous pointerawupdate)
8. pointermove for pointerId: 2 (should be diff from the second pointermove)

ccing: @smaug---- @mustaqahmed @daxpedda

[mustaqahmed]

Thanks for filing this issue!

Before we look into this, I would suggest a simplified view of the problem here because PointerLock affects only mouse pointers by definition, and that in practice no system exposes a non-primary mouse pointer (see the last "note" in Section 4.1.2). So I would suggest dropping the pointerId condition here and adding a line in the spec to emphasize that PointerLock is only about primary mouse pointers. What do you think?

[masayuki-nakano]

Well, honestly, I don't have concrete ideas how to use movementX and movementY of pointermove so that I'm not sure whether managing per pointerId is over action for the web API or expected by the API users of pointermove.

On the other hand, about movementX and movementY of pointerrawupdate, in my understanding, pointerrawupdate is useful for pen input handling and there is .persistentDeviceId. So, I guess that pointerrawupdate users may assume multiple pointer usages and it seems that movementX and movementY are useful to compute smooth curves at the path of moving a pen. Therefore, I think that should be managed per pointerId.

(In my understanding, movementX and movementY are also used without Pointer Lock. Therefore, it seems that managing per pointerId may be expected by the web developers.)

It's better to listen to the voice of web developers before considering the conclusion?