- From: Mark Nottingham <mnot@yahoo-inc.com>
- Date: Wed, 25 Jul 2007 15:26:24 -0700
- To: Stefanos Harhalakis <v13@priest.com>
- Cc: Henrik Nordstrom <henrik@henriknordstrom.net>, "Roy T. Fielding" <fielding@gbiv.com>, HTTP Working Group <ietf-http-wg@w3.org>
Because of the head-of-line problems which pipelining attempts to solve. HTTP connections being short-lived is largely a side effect of implementation techniques that more and more servers are overcoming. On 2007/07/25, at 1:14 PM, Stefanos Harhalakis wrote: > On Wednesday 25 July 2007, Mark Nottingham wrote: >> I've been wondering about this for a little while, and would very >> much like to see more work in this area. One of the interesting >> questions is how you'd layer HTTP-over-SCTP onto the existing Web; >> Upgrade doesn't quite meet the requirement (it's a whole new >> connection). Seems like one more piece of site-wide metadata to >> advertise... > > Why one would want to implement "HTTP over SCTP"? AFAIK SCTP is > not meant > for this kind of usage. HTTP connections are short lived and have a > strict > requirement for ordered data delivery. They also seem to better match > the 'stream' instead of the 'message' notion. AFAICS, SCTP should > be used for > long (or at least longer than HTTP) lived sessions that communicate > using > messages instead of a stream and possibly don't require ordered data > delivery. -- Mark Nottingham mnot@yahoo-inc.com
Received on Wednesday, 25 July 2007 22:26:44 UTC