Re: Note from a satisfied customer

[Dan Kegel <dank@alumni.caltech.edu>]

  At 05:57 PM 3/15/96 -0500, Bill Frezza wrote:
  >Interesting thing I've learned about that second B channel, at least in the
  >real world of Web surfing. Moving from 28.8 kbps to 64 kbps provides an
  >astounding improvement in throughput, but adding that second B channel does
  >just about nothing. Yup. It seems that most web servers top out at around
  >56 kbps anwyay, so all adding the B channel does is move the bottleneck to
  >the other end.  ... I had to actually create a contrived situation (four
  simultaneous
  >file downloads) to get the throughput up to 90 kbps.
  
  Good point.  Another thing that can get in the way of a faster line working
  well is TCP window size.   Many TCP stacks were written for 64k or lower speeds,
  and only support an 8 kilobyte window.  Going twice as fast means you
  need twice as big a window, and if your stack doesn't increase the window
  to 16 kilobytes or so, it might not seem much faster at 128k than 64k.
  (And ADSL might need a window size of 256K.)
  
  (The window is how much data can be sent before waiting for an ACK.  It
  compensates for network delays.)
  - Dan