Dear Ian, excuse me if I am bold, but I worked 19 years at IBM and 14 at
    teleprocessing dealing with all kinds of telecommunication systems and
    modems in several different communication protocols, I also developed
    some communication devices to IBM.  I am not challenging Horowitz, but,
    DTE asserts RTS (Request to SEND) when it wants to "SEND", not a single
    moment before, except if it is a full-duplex system, when the
    transmission and reception can happens simultaneously, then RTS still
    active all the time (can be programmed to be).

What you say is true of IBM and other "true half duplex" rs232 systems or
others fully obeying the rs232 spec.  This represents approximately 0.05% of
existing asynchronous communications equipment, nearly all of which is full
duplex (removing the need for "spec" RTS/CTS behavior.)


    In any case, RTS is not used (never) to means DTE is ready to receive.

Sure.  Except for about 2 billion modems, printers, terminal servers and
other async serial devices that implement "hardware flow control" using
RTS/CTS.  Asserting RTS/CTS means "ready to receive" (buffers not full), and
this will typically stop transmission of the other side at the hardware
level.  Sorry spec writers, you were overruled by expedience.  There is
no "hardware flow control" in rs232.  It was needed, it was added. (and it's
just like IBM to claim it doesn't exist :-)

BillW
cisco