BURATI@ulowell.CSNET.UUCP (07/25/86)
We are running both versions of telnet on our apollo net, talking to VAX 11/750's running Ultrix, an 11/780 running VMS and a MV10000 running DG/UX. Here's the problem, /usr/ucb/telnet talks to everybody fine, hangs up after a session fine... But, /com/telnet is a different story. It will call any of the other machines fine, work fine during the session and even close fine after a session to a unix machine. But to VMS, which is running EXCELAN telnet and ftp, when I log out, the /com/telnet starts spitting out errors "Remote machine reset connection" or something to that effect. Trying to escape from /com/telnet then becomes an adventure, Control-q won't work and any other key sequence just repeats the error "Remote machine reset connection"... If I try to sigp the process from another shell, it ignores it. A sigp -B from another shell causes any input to the locked window to be completely ignored, and the process disappears from the process list. But then the window can't be deleted (even control-n is ignored). Finally in desparation, I turned it into an icon to keep it out of the way and it turned into a graphics icon. When I tried to log off it was unable to stop all processes and I had to tell it to BLAST them. What's the major difference between the two telnet utilities (that question directed mostly towards any Apollo R&D's listening). I realize that the VMS telnet may be the one messing up /com/telnet, but there should still be a way to kill it without all this trouble. Michael Burati University of Lowell CS Dept UUCP: {apollo | wanginst | masscomp | datacube}!ulowell!burati CSNET: burati@ulowell.csnet ps: I'm not unhappy with the software, in fact except for this one problem, the net software has been perfect since we got the sr9.2.3 version (sr9.0 tcp software gave me nightmares).
Hans-Werner_Braun@UMich-MTS.MAILNET.UUCP (07/26/86)
If the Apollo implementation does not close a connection if it gets a reset from the other peer than it is broken. Note that Unix 4.3 will send you resets in some cases, too. I mentioned to Excelan that they set the RES bit instead of the FIN bit some while ago, and I think it is fixed in a newer version (at least for VMS, but I assume for their other implementations, too) except where it makes sense to RESet to avoid hanging connections. Even though I found some problems with the Excelan implementation I got the impression that Excelan is eager to listen (if one tells them) and to fix the problems.
rees@apollo.UUCP.UUCP (07/31/86)
Here's the problem, /usr/ucb/telnet talks to everybody fine, hangs up after a session fine... But, /com/telnet is a different story. It will call any of the other machines fine, work fine during the session and even close fine after a session to a unix machine. But to VMS, which is running EXCELAN telnet and ftp, when I log out, the /com/telnet starts spitting out errors "Remote machine reset connection" or something to that effect. The Excelan telnet server does indeed reset the connection instead of closing it. This is a bug on their side. They are aware of the problem and have promised to fix it. You could argue about what the right thing to do is when the connection is reset. I assert that we do a reasonable thing -- tell the loser that the connection was reset, then await further instructions. Probably we should put the user into telnet command mode instead of staying in conversation mode, since the connection is unusable at that point. The telnet protocol spec is no help on this matter. Most users seem to want us to just exit telnet, the same as bsd4.2 telnet does. The manual says how to close a telnet connection. I won't spoil your fun by telling you here, but I will say that if you blast telnet (or any process, for that matter), you get what you deserve -- utter chaos.