- From: Patrick McManus <mcmanus@ducksong.com>
- Date: Tue, 21 Jul 2009 20:22:14 -0400
- To: Adrien de Croy <adrien@qbik.com>
- CC: Jamie Lokier <jamie@shareable.org>, Mark Nottingham <mnot@mnot.net>, Henrik Nordstrom <henrik@henriknordstrom.net>, HTTP Working Group <ietf-http-wg@w3.org>
On 7/21/09 6:30 PM, Adrien de Croy wrote: > > There are many reasons in here why it's not commercially desirable to > open many many connections. I've always wondered why this is in the specification at all as a limit. As we all know, and this thread has shown, there are pros and cons to different approaches and different contexts change that math. It really has very limited (if any) relevance to interop, and the existing "2-per" rule is widely ignored (imo both because it is unreasonable for many current contexts and because it really has little impact on interop). I would favor just dropping the requirement and replacing it with some non-normative language briefly discussing the implications of high or low levels of parallelism vs pipelining. I know I've ignored this requirement in the past without regret.
Received on Wednesday, 22 July 2009 00:22:51 UTC