On Fri, 1 Nov 2002, Peter Seebach wrote:

> In message <Pine.LNX.4.21.0211011646140.13900-100000@Tempo.Update.UU.SE>, Johnn
> y Billquist writes:
> >It's a problem, because at the same time, you want to have this as close
> >to the terminal as possible, since you have a short time between sending
> >the xoff until it is reacted upon, during which you will get additional
> >characters. Because of this, terminals using software flow control must be
> >able to handle a few more characters coming in after the xoff have been
> >sent. On "modern" terminals you can select at which point the xoff is
> >sent, so that you better can handle slow reaction to the xoff.
> 
> Ah-hah!

I hope it helps. Obviously, you could have quite a lot of characters
queued up in ssh and the tcp/ip stack, so that while the xoff sill stop
any more output from the source, the buffers could hold several thousand
bytes still pouring out to your terminal. Buffered networks, along with
delays, really play havoc with xon/xoff. To work properly, it should
probably be handled locally...
But that will mean that ^S won't work in Emacs... :-)
Which is why I (enable-flow-control-on 
"vt52" "vt100" "vt200" "vt300" "vt400") in my .emacs. That rebinds ^S to
^\

	Johnny

Johnny Billquist                  || "I'm on a bus
                                  ||  on a psychedelic trip
email: bqt@update.uu.se           ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol