hedrick@athos.rutgers.edu (Charles Hedrick) (09/04/88)
Several people have asked recently about bringing up the KA9Q TCP/IP implementation under Microport Unix. I have diffs for doing it under SV/AT. I posted these before, so I think it would be better for me to mail them to people who are interested. (The file is 28Kbytes.) In my opinion, SLIP is just a toy at 2400 baud or less. (It still might be useful to help get experience with TCP/IP.) At 9600 baud it seems like it might be a fairly useful thing. In any case, note that in order to use SLIP, you have to find a system that is on an IP network and that is prepared to allow you to connect to it with SLIP. Just for the record, here are the things for SV/AT that I keep around. These have all been posted to this group, but as far as I know there is no archive site for this group. (I used to cc Microport on these things, hoping they'd put a copy of their bbs, but nothing seemed to come of that.) In general I send my changes back to the program author, so it's quite possible that in many cases these changes are not needed with the latest versions of the software. (This is not true with ksh, at least. Dave Korn considers the System V sxt code impossibly buggy, and is now supports job control only under Berkeley Unix or System V with Berkeley code merged in. Thus he has not accepted my diffs, and may even have removed the code that the diffs fix. That would be a shame, since job control really does work well on SV/AT.) ka9q - diffs to make it work under SV/AT, including workaround for a compiler bug and fixes for various problems with the way ka9q uses System V I/O . kermit - diff to make long packet mode work. This increases tranfer efficiency significantly. (This has been reported to Columbia and may be in the current release.) ksh - diffs to allow ksh to use the sxt device to emulate Berkeley job control. Note that ksh itself is an ATT product, although relatively inexpensive. So you have to get that source from the ATT toolchest, or convince Microport to release my version of ksh. (Job control is the ability to suspend a job that you are currently talking to, and come back to it later, and to move jobs between foreground and background. So you can start a job, then decide you want it to proceed as if you have run it with "&", then if it needs terminal I/O bring it back into foreground status to interact with it. It's probably the most important feature of Berkeley csh that is missing from sh (and most System V versions of csh -- System V people generally don't like csh because the copy they've got has all the interesting features left out.) malloc - a large-model malloc that (1) works [unlike Microport's], and (2) takes special care to minimize the number of times it does brk or sbrk calls. This speeds up programs that use malloc by a very dramatic amount. E.g. loading files into microemacs is much faster when it's built with this version. (The reason is the sbrk typically causes your program to be swapped. This makes it a very expensive system call. This optimization may be important only on machines with less than 2M of memory.) mg (micro Gnu Emacs) - changes to make it do file name and command completion and ? sc (spreadsheet calculator) - changes to make it work under SV/AT xlisp - minor changes to make it work with SV/AT