🚀 go-pugleaf

We get corrupted data when using TCP via ethernet then an async line.
The configuration is 
	|S|---ethernet---|G|---async-line---|D|
S and D are source and destination hosts, G is the gateway.
The async line runs at 9.6Kbaud with CRC and a MTU of 400
(i.e. the gateway fragements).
It happens with as few as 2 connections over the async line.
The corrupted data is usually an IP header (with the 0x45 version/hdrlen byte),
it occurs on 1K boundries in rcp'd files.

Has anyone else seen this, and do you have a fix?

	Eric Gisin, egisin@watmath.uucp, egisin@waterloo.{csnet|edu}