*BSD News Article 72785


Return to BSD News archive

Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!munnari.OZ.AU!news.mel.connect.com.au!news.mira.net.au!inquo!news.seinf.abb.se!nooft.abb.no!Norway.EU.net!nntp.uio.no!news.cais.net!world1.bawave.com!newsfeed.internetmci.com!in2.uu.net!in-news.erinet.com!news
From: geosh@erinet.com (Geosh Fathauer)
Newsgroups: demon.ip.support,demon.tech.unix,comp.unix.bsd.freebsd.misc
Subject: Re: Batch FTP and Web Pages
Date: 3 Jul 1996 17:38:23 -0400
Organization: EriNet Online 513 436-1700 (Voice)
Lines: 31
Message-ID: <4rep8f$bq@eri.erinet.com>
References: <31D4AA3A.BC0@www.play-hookey.com> <836073421snz@dsl.co.uk> <31D87436.7C7F@www.play-hookey.com> <836295557snz@dsl.co.uk> <4rcr6v$dh@anorak.coverform.lan>
NNTP-Posting-Host: eri.erinet.com
X-Newsreader: TIN [version 1.2 PL2]


Actually, the following is somewhat true. While I can't give the total
technical discussion, the gist is this. Many of the protocals involved
are optimized to take advantge of sustained connections; once the connection
is opened, both ends work together to get the fastest possible throughput.
Web browsing opens up an individual connection for every page and every
graphic. So, there are all the lag times to establish the routing, connections,
et cetera. Think of it like this: You with your 28.8 modem and your 
computer set at 115200; would you rather log on, browse a page
and get all the data and graphics at once, and log off, or have to
log on, get the http data, hang up, dial AGAIN, retrain to a proper
rate, get the next graphic, hang up, dial AGAIN, retrain to a proper
rate, get the next graphic, hang up, dial AGAIN... you get the picture. 
Yes, the net does and is actually working in a manner such as this.
The people who write web browsers are aware of this and are working
on protocals such that a web browser can request a full document,
and the server will serve up the whole bundle in one blow
on a single connection...

Geosh



: : Another problem with web surfing is that Netscape took a cynical decision
: : to launch parallel multiple sessions which are playing absolute havoc
: : with the RTT timeout adjustment algorithms of most TCP/IP packages: it
: : has been said that this is the single most contributory factor to the
: : slowing down of the 'net in the past year, with fully 40% of packets
: : being needlessly duplicated.

: I find that difficult to believe.  What do you mean by TCP/IP packages?