Yikes, that really seems like abuse of HTTP/2 semantics to me. My understanding has always been that server push in HTTP/2 is only intended to prime the client cache for requests you know they're about to (or likely to) make. [1]
> Pushed responses are always associated with an explicit request from the client.
I'd rather see additional frame types defined that allow for bidirectional communication (i.e. websockets 2.0) than overriding and conflating the PUSH_PROMISE frame type.
> Pushed responses are always associated with an explicit request from the client.
I'd rather see additional frame types defined that allow for bidirectional communication (i.e. websockets 2.0) than overriding and conflating the PUSH_PROMISE frame type.
[1] https://tools.ietf.org/html/rfc7540#section-8.2