The UA should be able to disallow speak() from autoplaying

[wicg]

The UA should be able to disallow speak() from autoplaying

[foolip]

Using the existing mechanism SGTM. Promises could be a later add-on that deals with all error cases.

Is there a linkable spec concept for "caller frame or any of its parents have receive user activation since their last load" ready use here? Is there some other spec which already uses this condition? (Sounds like something that would be defined in HTML, didn't go to look.)

[csharrison]

I don't think there is a linkable spec concept for this. This gets us into the guts of user activation which is not well specced at all at the moment. Fixing that is explored in whatwg/html#1903.

[foolip]

https://html.spec.whatwg.org/multipage/interaction.html#triggered-by-user-activation itself does need some work, that is true, but how about the "have receive user activation since their last load" aspect of this? @mounirlamouri, assuming this should match media elements and web audio, what's the right spec incantation?

[mounirlamouri]

For both specs, we do not require a user activation. We use a concept of "allowed to play" or "allowed to start" and we let the UA define when things are allowed. Because this is an intervention, UA have different rules and we decided that what mattered was discoverability of these rules by the website.

Regarding user activation, we have some folks working on a proper definition for it.

[foolip]

OK, so that is https://html.spec.whatwg.org/multipage/media.html#allowed-to-play.

@csharrison, referencing that concept directly won't work because it's defined as specific to media elements. Do you think an analogous definition for https://w3c.github.io/speech-api/#speechsynthesis would be best, or to generalize this concept and use it for all call sites? @mounirlamouri opinions?

[mounirlamouri]

For Web Audio, we used https://webaudio.github.io/web-audio-api/#allowed-to-start so it seems that there is precedent to re-define a concept.

[csharrison]

I've altered the title of this issue to be more general, and we can discuss the precise rules for allowing speak in #35.