On 21 August 2014 06:44, Mike Bishop <Michael.Bishop@microsoft.com> wrote:
> We're okay with the current text. An extension will specify in what
> states it's valid to send, so this only applies to unknown extensions on
> receive. Those will be discarded anyway -- anything beyond that will
> complicate implementations unnecessarily.
Mike,
The problem with deferring to extensions to " specify in what states it's
valid to send", that there are some undefined states
The state machine as defined by the document assumes that the arrival of a
trailer is atomic and switches state directly to half closed or closed.
But in reality there are extra states of quarter close, two quarter closed
and three quarter closed (see
https://github.com/http2/http2-spec/issues/484).
So if we wish to defer to extensions, then we should re-open 484 and define
all the states, otherwise it will be extremely difficult for any extension
to be correctly defined to work with closing streams.
cheers
--
Greg Wilkins <gregw@intalio.com>
http://eclipse.org/jetty HTTP, SPDY, Websocket server and client that scales
http://www.webtide.com advice and support for jetty and cometd.