New things in this distribution: Multiple Console Servers The client can now connect to several machines in succession when searching for a particluar console. The config file has changed to reflect this, replacing HOST with HOSTS, which is an array of strings, with a NULL string at the end. Note that only the client software needed to be changed to accomplish this. This introduces one bug, but I'm not likely to fix it: the mechanism for finding unique abbreviations can't work accross multiple console servers. For instance if you say "console tr" wanting to get "tree" on the second console server, you may get "tribe" on the first server, since it can't know what the second server has. The reason I won't fix it is that the cleanest solution is to stop accepting abbreviations, and I happen to like them. Carrier Detect The TIOCSSOFTCAR ioctl is used to turn off (ignore) carrier detect. The tty must be opened O_NDELAY for this to work. If you want to use carrier detect, set the variable "softcar" to zero. Serial Line Break I switched from using TIOCSBRK and TIOCCBRK, to TCSBRK, for generating the line break. We had a problem under SunOS 4.1.1, with a scsi terminal interface, where the old way didn't work (I can't imagine why the original method would fail - they _should_ be completely interchangeable). Paper Included A plain text copy of the LISA conference paper is included. I would have included a troff formatted copy, but their troff macros seem to be a bit non-standard. Alternative Sources There is now, or should soon be, another version of the console server available at j.cc.purdue.edu. For information, send mail to ksb@mentor.cc.purdue.edu (Kevin Braunsdorf). I'm not sure at this point of what all their changes are, but here's some of them: They use cbreak mode instead of raw, which give them more control over the connection, including things like backgrounding connections, but makes the connection less direct. They added various features to the command set. Changed the command line parsing (more standard, probably) They worked on a problem with silo overflows (we haven't had much of a problem with that here).