WebSpeech's SpeechRecognition onerror incorrectly reports ErrorEvent as event

[shareefalis]

See: https://developer.mozilla.org/en-US/docs/Web/API/SpeechRecognitionErrorEvent

[msftclas]

All CLA requirements met.

[shareefalis]

For a more robust implementation. SpeechRecognitionErrorEvent type has to be implemented but the spec is a draft and subject to changes. So ErrorEvent is a good enough implementation.
Reference: https://developer.mozilla.org/en-US/docs/Web/API/SpeechRecognitionErrorEvent

[saschanaz]

ErrorEvent has nothing to do with SpeechRecognitionErrorEvent. This is done because no browser implements the interface.

[sandersn]

It's OK, I guess, to use a different subtype that's "close enough". I'm not a huge fan, but I appreciate that default is easy even if it's not correct.

In the event that SpeechRecognitionErrorEvent is implemented, the PR that switches ErrorEvent to SpeechRecognitionErrorEvent will need to find and refer to this PR to explain why it's the right change.

[saschanaz]

@sandersn Despite the name, SpeechRecognitionErrorEvent has no filename, lineno, colno etc. from ErrorEvent and thus can confuse users more.

[sandersn]

@orta and I had a hard time deciding, but it seems like the choice is basically between a type with too many properties, ErrorEvent and a type with too few, Event. ErrorEvent is actually an unrelated subtype of Event, but that's mainly confusing to maintainers.

I went with the too-large type because it at least provides error and message even if filename/lineno/colno are also mistakenly present. If you feel strongly about this I will revert it.

[saschanaz]

That rationale sounds reasonable. Since the names of the extra properties strongly suggests they are irrelevant here, I don't oppose here. Thanks for the details behind the decision 👍