- From: Mike Bishop <Michael.Bishop@microsoft.com>
- Date: Tue, 2 Dec 2014 18:36:20 +0000
- To: HTTP Working Group <ietf-http-wg@w3.org>
- Message-ID: <BL2PR03MB132A10B901F7B8699C400A5877A0@BL2PR03MB132.namprd03.prod.outlook.com>
If I recall correctly, we chose not to define the semantics of PRIORITY from server to client. However, the state machine of what frames are acceptable in what states support the following: * Sending PRIORITY in "reserved (local)" (a server-only state) * Receiving PRIORITY in "reserved (remote)" (a client-only state) * Receiving PRIORITY in "half closed (local)" (a primarily-client state, unless the server responds before the request body has completed) * More generally, there's a note that PRIORITY can be sent or received in any stream state, presumably added as part of allowing prioritization of idle streams. Should we allow sending/receiving PRIORITY frames of undefined semantics? This seems like something to be cleaned up, either by prohibiting (my preference) or by including the semantics in the spec. And, as a side-note, the current text allows PUSH_PROMISE to be sent on idle streams, which I don't think was ever the intent. I submitted a pull request for this at https://github.com/http2/http2-spec/pull/665/files#diff-8894168382f6487e5e38c4306e613a88, because I think it's just an outright typo, but mentioning it on-list in case anyone disagrees.
Received on Tuesday, 2 December 2014 18:36:50 UTC