00:39
<Derrick>
Hey
07:23
<Andreu Botella>
I would prefer the final semantics are worked out ahead of time, as part of the "initial rollout".
That would mean analyzing essentially the whole set of events across the whole web platform, and would probably mean the whole implicit context propagation would need to be implemented from the beginning in browsers, which might result in interoperability bugs because the exposed surface is huge and it might be really easy for interoperability bugs to creep in
07:25
<Andreu Botella>
Also, the set of events we're considering for the initial rollout are ones for which we have use cases at this point, but there might be other use cases that haven't been brought up yet
07:26
<Andreu Botella>
One common thing with AsyncContext is that since much of the goal is for third-party code to be able to pass state seamlessly across first-party code, it's hard to know which events the context should propagate through, because the first-parties are not involved in the conversation
07:30
<Domenic>
I mean, saying people might implement it wrong and therefore we should implement it piecemeal is not generally how we do things on the web platform. We can't trade hypothetical interoperability bugs now for definite compatibility bugs later.
11:49
<Andreu Botella>
I am quite confident that there will be interoperability bugs because we're making things observable that so far have been unobservable
11:49
<Andreu Botella>
but fair enough
13:11
<annevk>
I guess what you're saying is that this new feature might build upon some existing infrastructure that might not match the envisioned behavior? But that's really an argument for writing sufficient tests. Not for delivering only part of the feature.