On Fri, 17 Feb 1995, John Vinopal wrote:

> This last line is quite in error.  Waiting for 1/2 of a 1024 byte
> packet is 512 / cps.  On a 9600 line, cps is 1200.  So that makes
> a 426ms wait for a 1024 MTU at 1200 cps (9600).

Faulty assumptions here lead to your erroneous times.  You forgot that
each character is preceded by a stop bit and followed by one or two stop
bits (but for most PC hardware and 8 data bits, only 1 stop bit).  In 
other words, 9600/(1+8+1) = 960, and 1/2 * 1024 / 960 is indeed 533ms.

--scott