Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!ucsd!ucbvax!FTP.COM!jbvb
From: jbvb@FTP.COM (James B. Van Bokkelen)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: TCPIP ping problem
Message-ID: <9012190142.AA07024@ftp.com>
Date: 19 Dec 90 01:42:58 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Reply-To: jbvb@ftp.com
Organization: The Internet
Lines: 25


    Machine A            Machine B                   Result 
    ==========           ==========                  =======
                         ping A                      ARP fails. Host unreach..

Tell me what INET DEBUG says on each machine at this point.  In particular,
how many packets were received, and any non-zero error counts.

    It seems to me that machine A must be caching the ethernet address of B
    after B does "ping A" otherwise its own ping would fail.  If so, why do
    the caches look empty?  Any help would be gratefully received.
    
B broadcasts an ARP request.  A receives it, sends an ARP reply, and cleverly
(courtesy of PCIP) caches B's address, assuming that it will need it soon.
I seem to recall that this optimization is suggested in RFC 826...

My first guess is that the Ethernet card in B is defective, and can't
see broadcast packets.  Compare the number of "unknown types" in INET
DEBUG and the number of "UDP no port listening" errors on A and B.
Most LAN OS file servers and Unix boxes produce some sort of periodic
broadcast, which you'll see in these counts (A & B would be roughly 
equal if both boards were working).

James B. VanBokkelen		26 Princess St., Wakefield, MA  01880
FTP Software Inc.		voice: (617) 246-0900  fax: (617) 246-0901