Client postmessage uses MessageEvent

[gterzian]

See https://w3c.github.io/ServiceWorker/#dom-client-postmessage-message-options

Should it not be an ExtendableMessageEvent, as is used in https://w3c.github.io/ServiceWorker/#service-worker-postmessage ?

When using a MessageEvent, it's not clear how to turn the service worker source into a window proxy source.

[yoshisatoyanagisawa]

I assume that clients usually use MessageEvent for postMessage(). In what specific cases would a client (especially a Window client) need to extend its lifetime using the Extendable interface?

[gterzian]

My point is rather that the source of a message event is defined at https://html.spec.whatwg.org/multipage/comms.html#dom-messageevent-source, and it does not include a service worker as potential source. So either an extendable event should be used, with a service worker as source, or the html spec should be updated to include a potential service worker source.

In other words, the sw spec contains a type error.

[asutherland]

In the MessageEvent WebIDL MessageEventSource is defined to be allowed to include ServiceWorker and is the type of source.

readonly attribute MessageEventSource? source;
...
typedef (WindowProxy or MessagePort or ServiceWorker) MessageEventSource;

For the source definition it does seem like the "it represents" portion should indeed be expanded to explain the ServiceWorker case because it's at the very least misleading to not have it there.

Pedantically speaking, is the "It represents" normative? Like it does seem like The source attribute must return the value it was initialized to. is a stronger statement than the "It represents" that is failing to be exhaustive. And the text in the ServiceWorker Client.postMessage algorithm does explicitly initialize it:

Let source be the result of getting the service worker object that represents contextObject’s relevant global object’s service worker in targetClient.
...
Dispatch an event named message at destination, using MessageEvent, with its origin initialized to origin, the source attribute initialized to source, the data attribute initialized to messageClone, and the ports attribute initialized to newPorts.

Which is all to say, it seems like the ServiceWorker spec is maybe fine but the HTML spec should have that piece enhanced?

[gterzian]

Thanks for pointing out the idl, I hadn't double checked it in the spec, only in the repo (Servo).

Yes I think the solution then could be just updating the wording in https://html.spec.whatwg.org/multipage/comms.html#dom-messageevent-source