2

I'm interested to know if anyone has worked on a website where they need to cater for users with vastly different bandwidths? I'd like to tailor a site to different user groups

1) Low Bandwidth ( <56kbps dial up)
Low Res Images - no javascript framework, no background images

2) Medium Bandwidth ( ~256kbps dial up)
Medium res images, smaller background images, small amount of jquery

3) High Bandwidth ( > 1Mbps)
Full site

Are there any techniques that I could use to test bandwidth speed, then maybe set a cookie to remember (to stop page running test every single time), and tailor included css/javascript and image urls depending on the result?

skaffman
  • 398,947
  • 96
  • 818
  • 769
calumbrodie
  • 4,722
  • 5
  • 35
  • 63

1 Answers1

1

At my previous company (a WiMAX ISP), I wrote a speed test for customers so that they could test the speed of their connection to our network. I ended up with a custom web server hacked together in Perl, that served static pages in a sub-domain.

The speed test itself was a JavaScript tool that sent a randomly created(*) string to a special address on the server, which made a mark when the transmission began, then when the string was completely received, and then sent an equally random string back to the client, marking the end of this operation. Now the server had three timestamps (start receive -> end receive == start send -> end send), calculating upstream and downstream speed and sending these as the last line of text in its response. The JavaScript part extracted the last line and knew the connection speed.

If you use a standard hoster, chances are you can't just implement your own server, though...

(*) Random string generated so that some routers could not taint the measured data due to compression of the the data stream

  • Thanks - your method would work but I'm guessing takes a while to work out the speed?? What size strings were you sending? I also assume that you were testing broadband speeds (is this correct?). I'm looking for something that can be seamlessly integrated on a site (non obtrusive). – calumbrodie Jan 25 '11 at 20:33
  • Yup, broadband. The trick was to have the server measure the time it took to download the page that contained the script to get an approximate downstream speed, upon which the number of bytes to use in the real speed test was determined (from 10k to 5M). Basically the more accuracy you need, the more data you need to send/receive. For an approximation, some k would be enough. –  Jan 26 '11 at 06:49
  • Thanks. Might be an idea to assume low bandwidth. Perform the test (in the background with ajax - using as little code as possible) - then reload the page at higher settings if bandwidth permits... – calumbrodie Jan 26 '11 at 12:10
  • Decided that a speed test would be counter productive as it takes too much bandwidth to get anything worthwhile. Will just offer a link to hi-res/lo-res version of the site. – calumbrodie Feb 25 '11 at 09:57