- From: Amos Jeffries <squid3@treenet.co.nz>
- Date: Wed, 30 Jan 2019 17:29:50 +1300
- To: ietf-http-wg@w3.org
On 30/01/19 3:56 pm, Mike Bishop wrote: > Cliff Notes of extensions: > > * Optional-to-understand frames can be sent without negotiation; > optionally also negotiate to discover there’s no point in sending > more of them after the connection is established. > * Mandatory-to-understand frames need to be negotiated first > > > > I could imagine a model in which the client sends priority in all > schemes it supports in the first flight, then picks one of them to > continue with once it sees which ones the server supports. That makes > the overall state on the server somewhat fluid, however. > For an optional extension the server which does not implement would treat any first-flight priorities in the base specification manner. One that does can start using the clients preferred way immediately. So from the client perspective it can be predictable - albeit more memory state to manage until the (non-)use is confirmed by arrival of the initial server SETTINGS frame. In theory, if one wants to put in the planning work ahead of time the weight field values might be made dual-purpose such that re-assigning early flight streams with PRIORITY frames is not necessary regardless of chosen interpretation of those 8 bits. Amos
Received on Wednesday, 30 January 2019 04:30:33 UTC