Taking a slightly different tack here, but the issue is with a single HEADERS+CONTINUATION+... block, right? I may need to re-read the draft, but I don't recall anything in there that explicitly forbids anyone sending HEADERS[EH], HEADERS[EH], DATA[ES] on a single stream. If such a restriction exists in the transport/framing layer, it should probably documented (more clearly, if it is already). Or, maybe better, it should be eliminated, along with the necessity for CONTINUATION frames. As far as I can see, that would only affect the entity who has to finally translate from HTTP2 frames to HTTP events (e.g. the client transport library buffering response headers so it can send a single "received initial headers" signal to the browser, if the browser expects a single signal). Please correct me if I'm mistaken here, and direct me to relevant text and discussion. -- Matthew Kerwin http://matthew.kerwin.net.au/
Received on Thursday, 29 May 2014 00:21:32 UTC