W3C home > Mailing lists > Public > www-talk@w3.org > March to April 2000

Re: Client header for connection speed info

From: Dmitry Beransky <dberansky@ucsd.edu>
Date: Thu, 02 Mar 2000 10:23:37 -0800
Message-Id: <4.3.0.20000302101056.00b0c210@biomail.ucsd.edu>
To: www-talk@w3.org
IMHO, this wouldn't be very useful.  For example, my usual transfer rate to 
Qualcomm's servers is 200-400KB/s, but when I visit sites in Russia, I'm 
lucky to get 6-7KB/s.  So what should my header say when I connect to your 
site?  Let's say I come to your site with a 1.5Mbs value in the 
header.  You start serving me high bandwidth content, then someone posts a 
link to your site on /. and before you know the slashdotters are 
oversaturating your site's bandwidth and you can't sustain the high 
transfer rates any longer.

The best way to deal with this issue is to maintain a running average of 
the bandwidth for each session on your server and adjust the content in 
real time accordingly.

Cheers
Dmitry

At 09:15 AM 3/2/00, Tam Freestone-Bayes wrote:
>Many sites and their users can benefit from the provision of more than one 
>http interface to a site - for example, the increasingly widespread 
>bandwidth discrepancy between standard modem and broadband connections may 
>encourage developers to produce both high- and low-bandwidth versions of a 
>site.
>
>Many sites implement this but rely on the user to manually select which 
>site they "prefer".
>
>If the site developer knew what the user's average download times were 
>upon their arrival, then this could be used to automatically determine 
>what level of information to display to the user before the first URL is 
>actually served.
>
>I suggest that a client side header (for example "Avg Speed: 38552" 
>representing bits per second, say, could be useful. It is the 
>responsibility of the client to calculate and maintain this figure based 
>on their average or maximum download speeds over the duration of a 
>browsing session.
>
>The header should obviously remain optional in any client implementation, 
>but the benefit of supplying it is high.
Received on Thursday, 2 March 2000 13:19:02 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 27 October 2010 18:14:24 GMT