tim@cmu-cs-k.ARPA (Tim Maroney) (08/23/85)
I am finishing an implementation of TCP/IP over Appletalk, the Macintosh local area network. The implementation is based on MIT's TCP/IP for the IBM PC. One problem remains. When TELNETing to a 4.1bsd VAX, it is not possible to initiate a close. The FIN segment gets sent out, a little extra data comes in from the VAX (and is ACKED by the Mac), and then nothing until the VAX resets the connection some seconds later. The VAX never sends out its FIN segment. I have gone over packet logs for hours, vainly looking for some deviation from the RFC in the behavior of my TCP, but there is none; I am matching the "Normal Close Sequence" diagram perfectly. Furthermore, when I TELNET to a TOPS system, the close sequence works fine. I tested this with the original MIT IBM PC version of TELNET using a borrowed PC, with the exact same results. Eventual reset on the VAX, immediate close on TOPS. Yet I have also frequently TELNETed from TOPS-20 to a 4.1bsd VAX, and that works fine from open to snag-free close. So I am hesitant to point the finger at the VAX. Unfortunately, I have no way of monitoring the Ethernet to find out what the TOPS and the VAX are saying to each other. Does anyone know why the MIT PC TCP/IP does not close properly with the VAX 4.1bsd TCP? (Incidentally, TFTP and TELNET for the Mac over Seagate will be appearing in net.sources.mac as soon as I get this one last problem squashed or at least figured out....) -=- Tim Maroney, Carnegie-Mellon University, Networking ARPA: Tim.Maroney@CMU-CS-K uucp: seismo!cmu-cs-k!tim CompuServe: 74176,1360 audio: shout "Hey, Tim!"
kevin@harvard.ARPA (Kevin Crowston) (08/27/85)
I've had similar experiences with TELNET between a Xerox 1108 and a BSD 4.2 VAX. After the session is over, the TCP processes seem to hang around forever, waiting for a CLOSE, or something similar. Perhaps we should compare notes to see if the effects we're seeing are the same. Kevin Crowston MIT Sloan School of Management kevin%mit-sloan.mit.edu@mit-mc.arpa kevin@harvard.arpa -- Kevin Crowston UUCP: {seismo,ut-sally}!harvard!kevin MIT Sloan School of Management ARPA: kevin@harvard.ARPA