On 27 Mar 2014, at 03:26, Martin Thomson <martin.thomson@gmail.com> wrote: > On 26 March 2014 14:57, Harald Alvestrand <harald@alvestrand.no> wrote: >> In fact, you might want to read up on HTTP/2.0 - especially why it's >> neccessary, and why it uses a single congestion-control connection between >> each client/server pair, not multiple ones. > > ...and why per-stream flow control (i.e., back pressure, which is not > much related to congestion control) is critical to the operation of > the protocol. > > I think that we should be asking the IETF to fix SCTP for us (as > opposed to patching its deficiences over the top). We thought about this in the past, but it is non-trivial. This is also one of the reasons why (in the socket API) there are no sockets per file descriptor or you can't read from a particular stream. So I would like to understand first what API you want and think about how to make that work from a protocol perspective second. Simple start/stop semantics can be realised by transferring reliably corresponding messages. Best regards Michael > >Received on Thursday, 27 March 2014 13:44:52 UTC
This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:17:55 UTC