large number of TX packets on ftp

Shawn Ostermann sdo@picard.cs.OhioU.Edu
Wed Jul 29 10:18:32 1998

> David SeSimone wrote:
> > Peter Stein <> wrote:
> >
> > Shouldn't the receiving end of a ftp session only attempt to send
> > something if a problem occurs?
> Normal TCP/IP implementations send an ACK(nowledgement) back to the
> originator to notify that packets were successfully received and do not
> need to be retransmitted.
> A good implementation will only ACK every so often, and not for every
> single packet.  This is determined by the window size.

To be exactly correct, TCP will either send back a very short ACK for
EACH data packet (segment) or will (using delayed ACKs) send back a
short ACK for every PAIR of data packets.  Sending fewer than that is
a violation of the standard, although they sometimes get lost
(normally without dire consequence).  It's the absense of these ACKs
that signals problems (TCP uses "positive" ACKs, some other protocols
use "negative" ACKS, giving the behavior that Peter assumed).

For a very slow connection (like a modem), you can see this behavior
on your modem LEDs; once you understand what TCP is doing, the lights
are a great indicator of what's happening on your connections (which
is why I dislike internal modems!).  For a fast connection, you can
sometimes see the same behavior watching the intensity of the LEDs on
an ethernet hub.

   Dr. Shawn Ostermann  -  Assistant Professor  -  Ohio University
      140 Morton Hall, Ohio University, Athens, Ohio  45701-2979 -- FAX: (740)593-0406 -- Voice: (740)593-1242