[comp.sys.hp] CMU IP/TCP V6.3 problem: uVAX <-> HP <-> HLH Orion

keith@mva.cs.liv.ac.uk (10/03/88)

We have quite a major problem in getting HP 9000/300 equipment to talk
TCP/IP with a microVAX II running VMS V4.7 and CMU IP/TCP V6.3 over
ethernet.

Basically, FTP and TELNET connections don't seem to last very long - the
microVAX aborts with TCP Receive errors. Worse still, using PING on the HP
system to ping to the vax with packet sizes in excess of 800 (or so) characters
causes the IP_ACP on the microVAX to completely abort with an access violation. 

We have absolutely no problem communicating with an HLH Orion as long as IP
trailers are turned off and the Orion communicates quite happily with the HP
system. 

As the the HP system doesn't appear to support IP trailers, we assume that
there is a resource problem. Should something in INTERNET.CONFIG be tweaked?
The documentation isn't very helpful. Any assistance would be very welcome...

Keith

Keith Halewood
Janet:     KEITH@UK.AC.LIV.CS.MVA
Internet:  KEITH%MVA.CS.LIV.AC.UK@CUNYVM.CUNY.EDU
UUCP(Ugh): {wherever}!mcvax!ukc!mupsy!liv-cs!keith   

marius@rhi.hi.is (Marius Olafsson) (10/07/88)

From article <2969@mva.cs.liv.ac.uk>, by keith@mva.cs.liv.ac.uk:
> We have quite a major problem in getting HP 9000/300 equipment to talk
> TCP/IP with a microVAX II running VMS V4.7 and CMU IP/TCP V6.3 over
> ethernet.
> 
> Basically, FTP and TELNET connections don't seem to last very long - the
> microVAX aborts with TCP Receive errors. ..

The CMU-TEK implementation periodically sends so-called "inactivity 
probes" over its connections to ascertain that the connection is still
alive. It is our experience that HP-UX always aborts the connection 
upon receiving these probes. To quote from the CMU-TEK code:

	                           ............ Currently, we will send
	an unacceptable segment with SYN and ACK on with a bogus sequence
	number. According to the TCP spec, such a segment should either
	generate an ACK with the correct sequence numbers or should generate
	an RST if the connection does not exist.

The problem is that HP-UX always generates RST. It would be nice if someone
from HP could comment wether this is a case of different interpretation of
the TCP spec or a bug in HP-UX.

Anyway, there does not seem to be a way to turn this off in CMU-TEK, and
the only way we had to make the HP-UX/CMU-TEK combination behave was to
patch CMU-TEK to eliminate these "inactivity probes". Mail for details
if anyone is insterested.

Our config (11/780 VMS 4.7+CMU-TEK-6.3 HP-9000/840 HP-UX 1.2)

--
Marius Olafsson 		Internet: marius@rhi.hi.is
University of Iceland		Non-MX:   marius%rhi.hi.is@uunet.uu.net
				UUCP:     {mcvax,enea,uunet}!hafro!rhi!marius