Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!utgpu!water!watmath!clyde!cbosgd!ihnp4!ptsfa!ames!amdcad!amd!intelca!mipos3!omepd!uoregon!dboyes From: dboyes@uoregon.UUCP Newsgroups: comp.os.minix Subject: Re: Future Patches To Make Minix Usable in the Real World Message-ID: <191@uoregon.UUCP> Date: Tue, 9-Jun-87 19:33:50 EDT Article-I.D.: uoregon.191 Posted: Tue Jun 9 19:33:50 1987 Date-Received: Sat, 13-Jun-87 04:48:17 EDT References: <114@jc3b21.UUCP> Reply-To: dboyes@drizzle.UUCP (David Boyes) Organization: University of Oregon, Computer Science, Eugene OR Lines: 123 In article <114@jc3b21.UUCP> gordon@jc3b21.UUCP (R. Gordon Price) writes: >I am not trying to insult Mr. Tanenbaum and MINIX. I think MINIX is the >greatest educational "UNIX TOY" around and can compliment the effort and >the method of distribution with the highest praise, however, I have >been here before with "VIRGINS" and have very little hair left to pull out. >I was there with "OS9" and "UNIFLEX" and a couple other efforts on micros >a few years back trying to get them to run in a semi-commercial atmosphere Whoa. I think you're missing the point here. Minix was never intended to be a production OS like OS9 or Uniflex. It's an EXAMPLE (admittedly one jim-hell of an example) intended to be something that students studying OS design can look at, tweak, fix, break, fix again, improve, etc., all while gaining a darn good understanding of how to design & implement OS code. Andy T. never claimed that Minix was anything but an educational tool. >and this whole MINIX process is like "DEJA-VU" to me. There are some glaring >problems with using MINIX as even a toy development system that have stopped It's not a development system -- it's an example intended to accompany a course in OS design -- it'd be useful in a compilers course, too. See above. >(1) the C compiler and MINIX upchucks floating point numbers at run time. > the documentation coyly warns of this. God, how do you write any kind > of applications even for fun only without real numbers?? HELP!! Simple enough. Most (more than 65%) of the applications written in the "real world" don't need floating point math, or can be written to minimize the use of such operations with a little thought. Besides, it's really in keeping with the educational orientation of Minix -- you need floating point, fine, go write it. You have source -- go for it. >(2) The boot operation leaves me wondering why a hard disk was not used > right off the top. Who would ever try to run UNIX on floppies?? Matter of cost. Educational institutions (and the rest of us underpaid slobs out here) are not exactly loaded with cash these days. LOTS of people have floppy-only systems. Visit your local state university some time -- I'd wager that a vast majority of the systems don't have fixed disks. Minix will run on floppies in order to minimize the cost of the equipment required to run it. >(3) There is no multi-user capabilities provided as I read it. I only > need one for testing purposes. I know the processor is brain dead and > that is the basic problem. Has anyone written a /dev/tty02 or higher > that works?? Are you serious? 2 or more users on a 4.77 MHz 8088? Remember that not everyone has AT-class machines, friend. Someone posted a /dev/tty02 driver not too long ago, and again, you have source -- go to it. >(4) The / directory is a flying joke. This must be because of the need to > use 2 floppy drives instead of a hard disk. If you accidentally copy > a large directory into slash / you are up for some entertainment. Agreed. However, Minix is very like Sys 7 Unix on the old PDP-11's, and they handled / exactly like Minix does. Blame this one on the model, not Minix. 2.9 BSD has this problem, too. >(5) The ram disk uses up too much memory that could be used by other tasks. > Can this be killed and sent to disk where it belongs?? Care to swap disks every time you want to do a cat foo? I don't. Besides, you can change the size of the RAM disk to anything you want and recompile the kernel. To quote an old beer commercial, "Fix it yourself, ape man!" >(6) NO SWAPPING IMPLEMENTED IS A HORROR!! What can I say? Sys 7 didn't swap either -- this ain't vmunix. Plus, you have to take into consideration hardware constraints -- you can't swap to floppies -- you yourself complained about how slow floppies are. Memory management code would have been an improvement, however I don't know of a memory management unit for the 8088, do you? Requiring memory management hardware would have made it next to impossible for many of use to afford to run Minix at all. Just as a quick comparison, how many private individuals do you kinow that have sprung the $150 or so for an 8087? Not many, I'd bet. >(7) MINIX does not appear to work with more than 640k of ram. Can this be > expanded?? Well, you have a 1 M address space in the base PC, much of which is occupied by IBM ROMs that are pretty tough to deselect. If you have a 286 based environment you could use protected mode memory, but Intel's brain-dead way of accessing it almost defeats the purpose. You could use LIM or EMS boards, but then the cost issue comes up again. Even Aboveboard clones aren't cheap. YCAHTKTDAYW. (You Can Always Hack The Kernel To Do Anything You Want) >(9) FSCK should be renamed "Fragment,Crash,Collate,Kill". If you use an > entire hard disk partition for MINIX which works fine by the way, FSCK > will screw it up because FSCK is expecting partitions 1-4 and not > partition 0.. If I ignore running FSCK I have no problems. A patch for this was posted early on. > Hopefully some of my >complaints have been considered for solutions pending by the author or >users. I would love to see MINIX reported to a 68000 or higher where it Somoene already has a NS32K port, and Since You Have Source, a 68K port wouldn't be all that tough. Go to it. In a lot of ways, you're trying to make something out of Minix that it wasn't intended to be. If you want a production OS, go buy SCO Xenix V.3 or Microport Xenix -- they're intended to be what you're talking about. Minix is an educational gadget and a grand example -- you could consider it a problem set in OS design. It's cheap, requires minimal hardware, and is exceptionally well documented (do you get anything like The Book with your AT&T source license?). Besides, for $75, I'm willing to do a little hacking to add features; after all it's saving me buyin an AT($2k), a 40 or 50 meg hard disk (~$800) and Xenix ($1200). > >R. Gordon Price >Century Computer Systems Inc. -- David Boyes ARPA: 556%OREGON1.BITNET@WISCVM.WISC.EDU Systems Division BITNET: 556@OREGON1 University of Oregon Computing Center UUCP: dboyes@uoregon.UUCP