Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!purdue!decwrl!ucbvax!ucsfcgl!cgl.ucsf.edu!tef
From: tef@cgl.ucsf.edu (Thomas Ferrin%CGL)
Newsgroups: comp.sys.proteon
Subject: Re: p4214 vs Cabletron multiport??
Message-ID: <11358@cgl.ucsf.EDU>
Date: 10 Feb 89 03:12:00 GMT
References: <8902091749.6713@uhura.cc.rochester.edu> <8902091829.AA07872@sonne.tn.cornell.edu>
Sender: daemon@cgl.ucsf.edu
Reply-To: tef@socrates.ucsf.edu.UUCP (Thomas Ferrin)
Organization: UCSF Computer Graphics Lab
Lines: 12

One of the strangest problems I ever had concerned connecting up a new
Excelan multiport transceiver to a network via an old Interlan NT10
transceiver.  This was on a large network with many machines.  When the
MPT was attached to the xceiver *some* pairs of hosts (which were *not*
connected to the MPT) experienced extreme (80%) packet loss.  Other hosts
ran just fine.  The hosts connected to the Excelan MPT ran fine.  Attaching
a Sniffer to the net showed that all of the missing packets where there on
the wire, it was just some hosts conuldn't receive them.  There were no
CRC errors, no runt packets, etc.  A TDR showed nothing unusual.  It took
days of troubleshooting to isolate the problem.  I do not know of a logical
explanation for this type of mysterious behavior, but I do know to avoid
mixing old and new spec equipment now.