Thanks -- but I am not an expert on this. Perhaps you can examine
rfc2001 first to make sure that my interpretation is correct (see
http://info.internet.isi.edu/in-notes/rfc/files/rfc2001.txt). I was
alerted to the possibility that Linux does not do this algorithm quite
right by Scott McMillan, who gave me the following explanation:
> From everything I've read, linux and other
> implementations really don't follow the "spirit" of the original algorithm
> introduced by Jacobsen in his 1988 SIGCOMM paper. What I mean is that
> linux only increments based upon individual acks coming in, NOT how many
> packets that ack is acking.
This might explain problems with small packets that we've seen. Large
packets work fine because at least every other one gets ACKed, but the
situation is entirely different with tiny packets.
I did not read Jacobson's paper, but rfc2001 is available online. I've
also found rfc2414 interesting, since the initial "cwnd" of 3 or 4 can
help avoid excessively slow start. Some results of this are reported in
rfc2415 and rfc2416. Of course, the essential TCP is presented in
rfc793, with further clarifications in rfc1122 and high performance
extensions in rfc1323.
Sincerely,
Josip
-- Dr. Josip Loncaric, Senior Staff Scientist mailto:josip@icase.edu ICASE, Mail Stop 132C http://www.icase.edu/~josip/ NASA Langley Research Center mailto:j.loncaric@larc.nasa.gov Hampton, VA 23681-2199, USA Tel. +1 757 864-2192 Fax +1 757 864-6134- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/