Aucbvax.5906 fa.info-vax utcsrgv!utzoo!decvax!ucbvax!info-vax Fri Jan 22 02:55:03 1982 Serious, VMS vs. UNIX? >From GJC@MIT-MC Fri Jan 22 02:49:56 1982 GJC@MIT-MC 01/22/82 05:32:35 Re: Serious, VMS vs. UNIX? To: INFO-VAX at MIT-MC CC: GJC at MIT-MC This is a long note, but I humbly submit that it is worth reading by those who want a hint of a perspective, crazy as it may seem, provided by the real experiences of the author at MIT. What about the people for whom "applications" programs are where it's at? I don't like to fool around with operating systems too much lately, so I enjoy the verbosity of commands and error messages, and the general reliability and idiot-proof nature of VMS. I can come in at night, bring down Unix, switch disks and floppies, and boot VMS, hack NIL all night on my VAX-780, (note: single-user 780's are nice), and even when unexpected things happen I can muddle through with the help of the thousands of possible error messages, crufty verbosity of DCL, and the 7 linear feet of manuals. *However*: During the day UNIX is better, because there are always experienced users and hackers around to fix things, and tell me about the trix and philosophy. There is a lot to be said for this. It is more than just a personal thing. [1] What happens when applications software wants to be a single program which provides the users programming environment? What happens when it gets moderately big? For example, right now the Lispmachine software binary image is about 24 megabytes of stuff. And that assumes the support of filecomputers to keep source codes. [For people with lots of Unix experience that is about 375 pdp-11 address spaces.] The demand is here for NIL to grow large, because people want Macsyma, they want Multics-style-Emacs, they want to use the lisp too, the translators and compilers, and they want all the debugging information loaded, and bugs fixed quickly. Also, lisp hackers want to develop the lisp at the same time, and other people want to work on things like new algebra systems. Lets face it, some of the people who want all this are "unsophisticated" about the limitations of standard computers. They are the same crazies who work on nuclear fusion devices not even known to be possible. They have little faith in computers though, they only go by what they have seen. When faced with programs this large, many years of philosophy, of hard experience, can go down the tubes. Nothing is easy, nothing is trivially transportable. What matters most is that you have enough power to bootstrap yourself to the next level of evolution. How do you invest this power? [2] Anyway, I'll tell you want happens. (Already it has happened at some NIL Macsyma sites). The system-managers come to you and say, "hey, why don't you store a lot of this crap in files." And you say, "what is this, a large address space system or not? How come I have to worry about the difference between a file and a page of memory?" Maybe my experience is warped just like the physics guys. The only large address space system's I've used have been Multics and the Lispmachine, and now NIL under VMS. So far there hasn't been any problem with VMS memory management, every once in while we just up the system-configuration parameters. Also, I came on the Multics scene late in the game. Years of fine tuning had been already done, new and faster hardware purchased, I never got to experience the "slow" multics, now of myth and legend. (I have heard some neat stories about when multics first ran, how it took 30 minutes to print its first ready message.) For me, multics is the place where the largest Macsyma programs run, beyond the scope of Vax's or (present) Lispmachines. {Color detail: Run in to order to generate programs to feed to CRAY-1's} Lets get back to the evolution issue. If something moves too quickly to the new order, or sets its sights too low, then growth rate may be stunted for a generation or more. But to procrastinate is dangerous too. NIL risked certain death by staying too long on the PDP-10. A cross-compiler with no address space left provides but a still-birth! Perhaps someday the details may make an interesting story. Summary: Points [1] & [2] are related. People are important in the sufficient power consideration, however, my "VMS vs. Unix advice:" "Beware of the exesses of reinvention, however comfortable it may be." "Beware of intelectual bagage, of false security in cheap transportability and software immortality." I'll put my faith in mud, not diamond, creativity can do more with mud. -GJC p.s. See the story of the Diamond-Trade in "The Alantic," very apropos. ----------------------------------------------------------------- gopher://quux.org/ conversion by John Goerzen of http://communication.ucsd.edu/A-News/ This Usenet Oldnews Archive article may be copied and distributed freely, provided: 1. There is no money collected for the text(s) of the articles. 2. The following notice remains appended to each copy: The Usenet Oldnews Archive: Compilation Copyright (C) 1981, 1996 Bruce Jones, Henry Spencer, David Wiseman.