changwan@ Yes, all typing and IME related actions are intentionally made non-cancelable.

I understand this is undesirable, but please find more details in https://github.com/w3c/input-events/pull/50

This is intended, I meant to fill you in on this.  I actually insisted beforeinput for regular compositions not be cancelable until we fully understand the consequences and have a clear use case, and the lack of cancelability is in "Level 1" spec we follow.  Adding cancelability would correspond to meeting level 2 spec: http://crbug.com/714791

See https://docs.google.com/document/d/1yPZEkHl_WOPjVeilZjE1XmlyjLCe-uS7THI0SboyrMM/edit#heading=h.b8dg6x6azemh for more reasons.

Ok, thanks for the quick replies.

I'm working on a mechanism to enable the cancellation, wish I could prove the feasibility soon.

FYI, the related use cases include:
1) https://bugs.chromium.org/p/chromium/issues/detail?id=692387
2) b/37444073

For 1), I'm not sure I fully understand the problem, but in my view cancelling beforeinput is basically the same thing as editing the textbox with JS to remove the effect of the composition.  In fact the latter case I think is easier to detect that we should send restartInput() than the former (because many use cases of beforeinput cancellation are actually to reinsert the input with different styling).  So I don't think it's necessarily a reason to support cancellation.

For 2), I was thinking we should probably add some specific input event for Enter and Backspace which seem to need special treatment, which might be sent on yukawa@'s special IME event proposal on https://b.corp.google.com/issues/37444073#comment10 .  Cancelling beforeinput which contains a lot of extra text than the Enter is not that helpful, I don't think.