speck@cit-vlsi (Don Speck) (08/13/85)
And I thought I was never going to see another UDA-50... A neighboring research group just took delivery of a vax-11/750. They decided that they should run 4.2bsd, and knowing nothing about it, they came to our department (Computer Science) for help in setting up their hardware and software, because we run 4.2. Putting the hardware together was easy and fun, and gave me ample opportunity to tell horror stories about RA81's and chide them on buying an RA60 in place of a tape drive (especially since it used up the last of their FY85 money). I'm worried that 4.2bsd may turn out to be a rude shock for them, yet I want to do the best I can to give them a fighting chance of managing it themselves. I assume that the most important question is, 4.2bsd from where? What I've always seen done before is cloning of an already stable machine, by simply restoring its latest dumps onto the new machine. "Oh, you're a new to this? Here, I'll fix you up *real good*". I'm afraid that our extensive local hacks would cause more confusion than they're worth. And /usr and /usr/src take up 400 Mb on our machine, which they can't afford. Also, they don't have licenses for most of it; just Sys V and VMS licenses. On the other hand, simply installing the 4.2bsd distribution tape worries me too, because it's so buggy. We have a particularly early tape, September 1983. Do more recent tapes from Berkeley have fewer bugs? I know that a vendor- supported Unix (Ultrix or Mt. Xinu) is what they really need, but they don't have enough money left for that. While I was copying the 4.2 tape for them, I figured I might at least apply some minimal set of patches to give them a chance. (NULL rusage ptr, FTRUNC security hole, ptrace of suid-text security hole, "panic munhash", tb par fault mask, load rev 98 microcode from `boot'; add support for DMF lineprinter port for their LN01; add NYU Deuna driver; substitute the MIT UDA-50 driver for the Berkeley one, and make the partitions conform to `diskpart's conventions). That still leaves about 490 out of 500 bugs for them to fix. Am I taking the right approach? How do the wizard's wizards deal with this? Don Speck speck@cit-vax.arpa