> I know that there have been implementations of RFC 1323 and 2018
> and 2001. Does any one have some view graphs showing the real
> benefits of these RFC implementations and the conditions it has
> been realized.
RFC 1323 provides larger windows that allow TCP to fully utilize a
long-delay pipe.
RFC 2018 provides the information about exactly which segments have
arrived and which have not. This allows a TCP sender to obey
congestion signals, but not back off to a 1 segment window when
multiple segments are lost.
RFC 2001 is obsolete. RFC 2581 is the current version. That RFC
really provides no performance gain to a single flow (OK, fast
retransmit and fast recovery are gains over taking timeouts).
Rather, that document outlines the congestion control mechanisms
that everyone must follow so the network does not collapse, causing
everyone to get little or no performance.
The two documents produced by the tcpsat WG have more verbose
descriptions and many pointers to the literature. I suggest reading
RFC 2488 and draft-ietf-tcpsat-res-issues-12.txt (currently in the
RFC Editor's queue).
Happy Holidays!
allman
--- http://roland.grc.nasa.gov/~mallman/
This archive was generated by hypermail 2b29 : Mon Feb 14 2000 - 16:14:58 EST