MSS/MTU/WINDOW settings

[crompton@NADC.NADC.NAVY.MIL (D. Crompton)]

Gareth,

    I am not an expert on this but we have sure been around and around
on this one here. One thing I see wrong is your MSS/WIN/MTU ratios. We
have a local 2400 baud subnet running datagram mode with the following
settings.  MTU=512, MSS=472, WIN=944. The MSS value is 40 less than the
MTU to allow for the TCP header. WIN is equal to or some multiple of MSS.
With these values I have seen Point to point (2 station) transfers of
close to 100/sec. Higher MTU values, like 1024, would probrobly help at
2400 baud but we found problems with some of our TNC's at this value.
TAPR TNC2 code will support greater than 512 byte KISS frames, some others
(like AEA) may not.

   The problem arises when a switch station within this group tries to
interface this net to the netrom world where the mtu must be 256 (or less).
The different MTU/MSS values cause unfull packets on the netrom side. On
our local net though we see no fragmentation and have very good results
using datagram mode. I assume you are using G1EMM and TCP timer linear.
Contrary to popular belief the world does not come to an end when not using
expoential backoff.

Doug