dhesi@bsu-cs.UUCP (Rahul Dhesi) (12/20/88)
(Moving this from comp.sys.ibm.pc to comp.unix.questions.) In article <4330119@hpindda.HP.COM> hardin@hpindda.HP.COM (John Hardin) writes: >While I agree with your prediction of >the role of Unix in the next few years, I can also see why there continue >to be propietary OSs. One reason is the inefficiencies of Unix. I am no >Unix kernel expert, so I don't pretend to know why, but I have seen that >a propietary OS can support many more time-sharing users than Unix when >both are run on the same hardware. Counterarguments are possible. For example: o Proprietary OSs often support more users than UNIX by restricting what these users can do. For example, they may not allow job control (only one active job) or make it inconvenient; they may not allow pipes (got to go via disk files, slowing things down, so it seems like more people are doing work but each person is getting less done); they may not have many of the UNIX utilities like mail, wc, grep, awk, etc., so you just don't do those things at all o Proprietary OSs probably don't have Usenet software, which may mean a lighter load on the machine. So you read news from your home machine, and post an article saying that the proprietary OS at work is more efficient :-) o Proprietary OSs may require a special terminal with local intelligence (e.g., IBM) thus offloading much work to a $2,000 terminal, rather than letting you use a $500 terminal for the same purpose. Thus, to support 100 active users, you might need an additional investment of $2000 * 100 = $200,000 (in addition to the cost of the central machine). The proprietary OS may appear more efficient if you only looked at the cost of the central machine and forgot about coaxial cables, front-ends, terminal multiplexors, intelligent terminals, and other paraphernalia that the proprietary OS may *require* rather than just *allow* o UNIX is less efficient than many proprietary OSs because it is written mostly in C. An individual vendor is welcome to rewrite more of the kernel and utilities in his own proprietary machine-dependent language and speed up UNIX, or a proprietary UNIX clone in such a machine-dependent language, without sacrificing compatibility with UNIX applications o It is possible to have a proprietary OS emulate most of the UNIX system calls, allowing both efficiency and portability o Hardware is cheap enough that the additional hardware cost to make a UNIX-based system as efficient as one with a proprietary OS is probably smaller than the additional cost in software and inconvenience of using a proprietary OS By the way, UNIX itself is a proprietary OS. Perhaps we should be talking about "portable" and "unportable" operating systems. -- Rahul Dhesi UUCP: <backbones>!{iuvax,pur-ee}!bsu-cs!dhesi
debra@alice.UUCP (Paul De Bra) (12/21/88)
In article <5254@bsu-cs.UUCP> dhesi@bsu-cs.UUCP (Rahul Dhesi) writes: >(Moving this from comp.sys.ibm.pc to comp.unix.questions.) > >In article <4330119@hpindda.HP.COM> hardin@hpindda.HP.COM (John Hardin) writes: >>While I agree with your prediction of >>the role of Unix in the next few years, I can also see why there continue >>to be propietary OSs. One reason is the inefficiencies of Unix. I am no >>Unix kernel expert, so I don't pretend to know why, but I have seen that >>a propietary OS can support many more time-sharing users than Unix when >>both are run on the same hardware. > >Counterarguments are possible. For example: > >o Proprietary OSs often support more users than UNIX by restricting > what these users can do. For example, they may not allow job > control (only one active job) or make it inconvenient; they may > not allow pipes (got to go via disk files, slowing things down, > so it seems like more people are doing work but each person > is getting less done); they may not have many of the UNIX > utilities like mail, wc, grep, awk, etc., so you just don't > do those things at all > I fully agree with Rahul. Several years ago I worked on a Vax 11/780 with VMS, mostly used by physicists running batch jobs and editing using a line- oriented editor (except for some who used edt on a vt100). Then my (beginning computer science) department started using (Gosling) Emacs, and started getting complaints that we were slowing down the machine. Later I developed a few m-lisp functions for calling the Vax Pascal compiler or Tex from Emacs, and automatically pop up a window with the error messages and jump to the right line in the main window... It took about 3 days before the physics department decided that this could not be allowed and installed a limit of 1 job per user. So this "highly efficient" operating system was only efficient if nobody tried to do anything like what one does all the time on Unix. Shortly thereafter we got a small 68000-based Unix box which worked much more efficiently for us, because we could get the same job done in less time. For similar reasons I prefer to run Unix (Xenix) on my PC instead of the supposedly more efficient MS-DOS, because with Unix I can get the job done in less time. Paul. -- ------------------------------------------------------ |debra@research.att.com | uunet!research!debra | ------------------------------------------------------
burzio@mmlai.UUCP (Tony Burzio) (12/21/88)
In article <5254@bsu-cs.UUCP>, dhesi@bsu-cs.UUCP (Rahul Dhesi) writes: > >While I agree with your prediction of > >the role of Unix in the next few years, I can also see why there continue > >to be propietary OSs. One reason is the inefficiencies of Unix. I am no > >Unix kernel expert, so I don't pretend to know why, but I have seen that > >a propietary OS can support many more time-sharing users than Unix when > >both are run on the same hardware. I continue to hear rumors that DEC is going to come up with a VMS shell for UNIX based machines and toss the old BLISS VMS code overboard. With their new 15 MIP machines, they should be able to mess up UNIX enough to allow emulation of a uVAXII :-) Imagine, RMS emulation in UNIX. If you think about it, VMS residing as a shell in UNIX will make Olsens' statement that VMS is/will be OSF compliant understandable... > By the way, UNIX itself is a proprietary OS. Perhaps we should be talking > about "portable" and "unportable" operating systems. Portable? Which version of UNIX are you using this morning? In all fairness, though, portability seems to be *ever* so close. Now if the marketing guys can keep their proprietary hands off... ********************************************************************* Tony Burzio * Ski Angel Fire NM... Martin Marietta Labs * mmlai!burzio@uunet.uu.net * *********************************************************************