Unix help requested

Mark Crispin MRC%PANDA at SUMEX-AIM.Stanford.EDU
Fri Dec 19 15:49:10 AEST 1986


     We are using a 4.2 Unix system as a means of interconnecting two
DEC-20's together for reasons I won't go into here (let's say that for
the same reason we are using a DEC-20 to interconnect two Unix systems!).
One of the DEC-20's (PANDA) calls up a local Telenet PAD and uses it to
connect to the Unix system (nttlab) which is located in Japan.  Once
PANDA sees nttlab's prompt, it then sends a login sequence for a special
account on nttlab that has telnet as the shell.  When it gets telnet's
prompt, it sends "open ntt-20" to open a TCP/IP connection to the other
DEC-20 (NTT-20).  It then sends a login sequence to NTT-20 which
eventually will get the application running on NTT-20.

     We have had problems where a job is left logged in on nttlab long
after both PANDA and NTT-20 have long forgotten about the connection.
Supposedly, the X.25 world should have hung up the "phone" on nttlab and
that should have nuked the nttlab job, but this doesn't already happen.
This is causing a lot of problems for nttlab's management.

     I have suggested that one way of reducing the likelihood of this
happening would be to set up the nttlab account so that it automatically
connects to NTT-20.  That is, PANDA only has to send the nttlab login
sequence and then wait for NTT-20's prompt.  That way, the number of
possible states could be reduced.  It should also be set up so that a
failed connection attempt, phone hangup, or ANY other error should
immediately log out the nttlab account.  There is no human talking to
this job, so ALL errors should result in a logout.

     I know this is possible because I have used such "pass through"
accounts in the past.  However, we don't know how to set this up.
Could one of the distinguished experts on this list tell us what to
do?  If possible, we want to eliminate ALL possible ways of talking
to nttlab once the job is logged in.

     Please reply to MRC%PANDA at SUMEX-AIM.Stanford.EDU and to
Nojima%NTT-20 at SUMEX-AIM.Stanford.EDU.

-- Mark --
-------



More information about the Comp.unix.wizards mailing list