hedrick@topaz.rutgers.edu.UUCP (05/26/87)
We are using a version of Sun 3.2 that has the 4.3BSD TCP code in it. (I believe it came from Seismo.) I have just diffed the code with the 4.3 distribution, and it looks very close. Our operator has been complaining that rdump's from a particular host don't work. The only thing different about this host is that it is at the other end of T1 line from the machine with the tape drive. I just tried a dump with tcpdump running on a Sun, watching the conversation. What happens is that suddenly the data just stops coming. The other side of the connection (i.e. from the machine with the tape drive) does the usual probes to make sure things are still alive. But nothing comes the other way. Netstat shows that there is more than a packetfull of data waiting in the output queue to be sent. I've tried it several times. In a couple of cases, from the sequence numbers it was obvious that a packet had been lost. In one case, this was not the case. (Though the ack could have been lost, and I wouldn't know.) One suspects that now and then the T1 line is losing a packet, and that for some reason no retransmission happens. Does anybody know of a problem in 4.3 TCP that sounds like this?