- From: John Franks <john@math.nwu.edu>
- Date: Fri, 16 Dec 1994 11:50:26 -0600 (CST)
- To: Dave Raggett <dsr@hplb.hpl.hp.com>
- Cc: john@math.nwu.edu, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
According to Dave Raggett: > > > Correct me if I am wrong, but I concluded from Spero's postings that > > nothing currently proposed including MGET, hold-open, or even HTTP-NG > > would improve (or even match?) the user's perceived performance > > currently given by Netscape. By this I mean the ellapsed time until > > the user can start reading *all the text* and the ellapsed time until > > the user can jump to a new link. > > HTTPng will win over the multiple connections used by Netscape, as the > latter lead to congestion problems, since the connections don't share > congestion info. In addition, using a single connection gets around the > slow start problem leading to better peformance. It will also be much > nicer to servers! > Perhaps. But by now we have all had extensive experience with Netscape and while there have been numerous theoretical objections to multiple connections I don't think servers have collapsed under the load. I suspect that most providers will simply see the increased load (and I don't think we really know how much it is increased) as the price of supplying a better product (from the consumer's point of view). As for the TCP slow start problem, I don't understand why a single slow start ala HTTP-NG is better for the UPP (user's perceived performance) than multiple slow starts in parallel ala Netscape. Even if HTTPng will win over multiple connections, I think it will have to be a dramatic win before client vendors will completely switch. > Simon explained all this in his notes. > Could you give a reference? Thanks. John Franks
Received on Friday, 16 December 1994 09:54:25 UTC