need for RTS/CTS at high speeds (was re: SCO RTS/CTS Setup)

[Greg Andrews]

In article <1991Apr8.173125.22219 at mccc.edu> pjh at mccc.edu (Pete Holsberg) writes:
>A computer magazine recently tested a number of communications programs
>at various speeds and published the results.  They said that at higher
>transfer rates (57.6K and 115.2K), they used RTS/CTS handshaking because
>(I'm paraphrasing) that's what you had to use when you had high speed
>modems.  This puzzled me because I've been using Trailblazers at their
>maximum speed for a couple of years now, with either no or XON/XOFF
>handshaking.  Could someone enlighten me?
>

(sounds like the PC Mag review - no comment)

RTS/CTS is a lot faster than XON/XOFF.  By that, I mean the computers can
respond to RTS/CTS signals faster than XON/XOFF characters.  This has two
effects:  (a) Data flow stops sooner preventing buffer overflow, and
(b) Because of (a), comm software can raise the flow control threshold
to allow more data to come in before making it stop.  The first prevents
errors that would slow down data transfer, and the second increases the
efficiency of the data flow by increasing the amount of time the data
flows vs the amount of time the data doesn't flow.

It's faster because dropping an RS232 signal line can be accomplished in
one or two bit periods (the time to transfer one bit at >38,400 bps).
It requires ten bit periods to send an XOFF and a certain amount of time
overhead while the transmitting computer figures out it got an XOFF and 
decides to stop transmitting.

More important, it's a transparent form of flow control.  Protocols like 
XMODEM and uucp can easily send XON and XOFF characters as part of the 
file data.  If those bytes are interpreted as start/stop commands instead 
of file data, the protocol will encounter errors.  If the channel uses 
RTS/CTS, the XON and XOFF bytes can pass transparently through and aren't 
misinterpreted.

If the data flow across your TrailBlazers doesn't use XON/XOFF then you
probably wouldn't have encountered the above problems with XON/XOFF.
Running uucp transfers with PEP-mode spoofing in the TrailBlazers doesn't
count because the modems automatically turn off their XON/XOFF control
when the transfer begins.


-- 
.------------------------------------------------------------------------.
|  Greg Andrews   |       UUCP: {apple,amdahl,claris}!netcom!gandrews    |
|                 |   Internet: gandrews at netcom.COM                      |
`------------------------------------------------------------------------'