- From: Jeff Pinner <jpinner@twitter.com>
- Date: Mon, 27 Jan 2014 11:40:39 -0800
- To: Martin Thomson <martin.thomson@gmail.com>
- Cc: HTTP Working Group <ietf-http-wg@w3.org>
Received on Monday, 27 January 2014 19:41:08 UTC
On Mon, Jan 27, 2014 at 11:34 AM, Martin Thomson <martin.thomson@gmail.com>wrote: > On 27 January 2014 10:59, Jeff Pinner <jpinner@twitter.com> wrote: > > assume independent clients so each connection becomes its own stream in > its > > own group > > That's a reasonable thing to do. The key questions are: > - do you want fate sharing to that degree? > - is there a meaningful distinction, priority-wise, that can be made > at this point? > - would the cost of a collision (two clients hitting the same > priority group) be noticeable? > > i capitulate that the cost of collision would probably not be noticeable nor would any distinction in priorities be necessarily meaningful > I'll also note that at 16m groups, each client gets, on average, 128 > requests before you need to make a new connection. That's not very > many. I'd have thought that you would be better off asking for more > stream identifiers (I'll note that we could use that reserved bit with > this scheme, but that only doubles the space). > for web browsing, sure, but for streaming services a single request per client is enough
Received on Monday, 27 January 2014 19:41:08 UTC