Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!utgpu!water!watnot!watmath!clyde!rutgers!husc6!bu-cs!bucsb.bu.edu!madd From: madd@bucsb.bu.edu.UUCP Newsgroups: comp.sys.ibm.pc Subject: Re: IBM new 'standard' Message-ID: <855@bucsb.bu.edu.UUCP> Date: Mon, 23-Mar-87 15:13:44 EST Article-I.D.: bucsb.855 Posted: Mon Mar 23 15:13:44 1987 Date-Received: Wed, 25-Mar-87 01:22:55 EST References: <701@imsvax.UUCP> Reply-To: madd@bucsb.bu.edu.UUCP (Jim "Jack" Frost) Organization: ODO (Organization for the Disorganization of Organization) Lines: 209 In article <701@imsvax.UUCP> ted@imsvax.UUCP (Ted Holden) writes: >Today's generation of cheap XT and AT clones are meeting the needs of >most computer users far better than they have ever been met before, with >hardware and software add-ons to do just about any kind of thing >imaginable doing nothing but getting cheaper and better by the day >and, while I see some areas of applicability for entirely new kinds of >machines, I believe most users would do best to stay right where they >are software-wise, perhaps migrating to stronger 386 based single-user >machines as they become available and inexpensive. > >What I mean is this: that over the next couple of years, numerous >slick-talking people are going to be coming up with all kinds >of reasons for abandoning our happy single-user DOS world and I, for >one, intend to tell all of them to go **** themselves. I can think of many, many reasons why we should switch to a multitasking DOS. Note that I did not say multi-user DOS -- there is no difference between the two. If you have a multitasking DOS, you immediately have multi-user DOS, although you might need to play with it a little bit to put two users on it. >The reasons for >replying thusly to anyone trying to foist a closed-architecture system >on us should be obvious even to the blind. Sure, but who's talking about closed architecture systems? I thought you were talking about operating systems. They're not the same, you know. >Then, there will be the people who claim that DOS isn't a real operating >system. Operating systems are the fruits of poverty, a necessary evil >for computers which have to serve many masters. Owning your own >computer obviates any need for real OSs even as acquiring great wealth >obviates any need for techniques of scrimping. Multi-TASKING can be >achieved in software on a single-user basis quite well without >re-designing your operating system e.g. Mystic Pascal e.g. the article >on concurrency using Turbo Pascal in the 3-86 issue of Dobb's, etc. and >you don't even need an AT or protected modes to do it. All this is true. However, IT IS NOT CONVENIENT TO DO IT THIS WAY. In fact, it's not even easy. Wouldn't life be easier on the developer if he could just fork tasks instead of having to do all the task switching himself? I've done it both ways. I like forking better. Also, the operating system method is usually faster than doing it in the application. >[..] >The April 14 issue of PC Magazine contains an article on the new DOS, >known generally as ADOS, which they refer to as ADOG, claiming that it >is quite slow and that a great deal of ordinary DOS software won't work >under it, particularly memory resident items such as sicekick or the >Mother Jones package. In fact, such items will only function as device >drivers, with all sorts of fun commencing with any attempt to install >two or more of them at once. And all so that I can allow other people >to use my conputer? Maybe IBM and Microsoft haven't gotten the message >of the PC revolution: NOBODY uses my computer except me, and THAT is >what the micro revolution is about. ADOS has to run slow. It's running on a terribly slow machine. Give them a break -- they cannot do miracles. You'll notice that those programs that perform multitasking ALL perform slowly. This is necessary, since you no longer have all of the CPU. As for memory resident software, who cares? The point being: with a multitasking operating system, you no longer need resident software. In BSD UNIX, you can just suspend the job you're working on and run someting else. If you want to go back to the first thing, you can suspend the new thing you started and pop back to the first. This is the essence of resident software, isn't it? To be able to switch between some tasks easily. But with this new twist, ANYTHING can become "resident" software. You might argue that "well, that's fine for BSD UNIX, but what if they don't build in a suspend function?" The answer to that is that it is possible to create a program that will simulate a suspend function, such as the "layers" program in System V UNIX. Any operating system that operates similar to UNIX will be able to have the same functions. If you look at MS-DOS, you can see the UNIX base to it, although they goofed in many areas. I would be surprised if they abandon the UNIX base, because much software depends on it. Therefore, it will be possible to make a suspend function if one is not originally implemented. (in fact, it is possible to kludge one now -- look at microsoft windows, for example, although this is far from optimal and doesn't always work) I thought that the micro revolution put the power of computers in the hands of the masses, not that it made people switch to single user systems. I can't afford a mainframe, but I can afford a PC. Hence I can get a computer now, when I couldn't just a few years ago. This, in my mind, is the result of the micro revolution. >What about virtual memory and the other wonders of protected mode? >Virtual memory was also an accomodation to poverty on multi-user >machines for which there was never enough memory. In 15 years of >programming, including applications in Cobol, statistics, maximum >likelihood modeling, software design and all sorts of things, I have >never actually SEEN any application which could not be programmed in >640K, the standard for PC and AT clones these days. Probably 70% of all >programs I have ever seen which do anything worthwhile and were well >written take less than 64K. True fanatics in need of really BIG arrays >will likely find more happiness using the Turbo Extender package with its >LIM and virtual arrays than they will looking elsewhere I can think of many things that require more than 640K to operate well. Wordprocessors with built-in spell checkers, for example. It's a pain to have to do disk look ups. Much faster to have it in RAM. But you need lots of RAM when your dictionary gets big enough, even using digitally-linked trees like they usually do. Notice that I said "operate well". Again, you can kludge a solution, using disk-swapping and the like, but it's far from optimal. Anything that needs to deal with large amounts of information will eventually need more than 640k. >Then there is the possibility of running UNIX on 286 and 386 machines. >Unix is my OS of choice for machines which for some legitimate reason >HAVE to be multi-user; I can't think of any reason why any sane person >should wish to run it on a single user computer, unless they like slow >performance, clunky and fragile file systems, lack of standardization, >and generally prefer 1976 software, such rubbish as TROFF, VI, ED, YACC >etc. to the software of the present. Make no mistake, the DOS software >world is casting a giant shadow right now, and UNIX is standing IN that >shadow. When the biggest and most critical selling point of all UNIX >systems is "runs DOS software", the conclusion to be drawn is obvious. What? WHAT? You've lost your mind here. [sorry, but this infuriates me] First, no machine is single-user. Only the operating system. It is true that many systems are designed with one user in mind, but you can make it have more than one with the right operating system. While many UNIX things are not new, they work fine. But you fail to see that there are always new programs being developed. Emacs, for example. Graphics stuff. You name it, it exists under UNIX, or could be made if you really wanted to. UNIX does not have a fragile file system. If you think so, you haven't looked closely. MS-DOS, on the other hand, DOES have a fragile file system. If your file system is munched on UNIX, it is possible to partially or completely recover it because of the way the file system is built. If you blow up a part of an MS-DOS file system, you're most likely out of luck. There is little redundancy in MS-DOS, so you'd have trouble rebuilding your filesystem (although CHKDSK does do a pretty good job, considering). My biggest objection here is that UNIX systems advertise that they run DOS software. Maybe so, but who said this is the biggest selling point? I wouldn't consider it to be. The biggest selling point would be that it runs UNIX software, with tens of thousands of programs available, debugged, supported, and even with source code. Be serious. PC's don't shadow UNIX systems. They just haven't been around enough, nor are they powerful enough. >Finally, there will be people telling us we need new kinds of machines >to keep up with the Macs, SUNS etc. vis a vis graphics applications. >Bullshit! All such things can be outperformed with DOS/PC compatible >components AT THE PRESENT, and at a fraction of the cost. EGA graphics >blow the MAC away right now, Metheus and several other people have >1024x1024 and 1280x1024 color boards now for which Autocad and DrHalo >drivers exist etc. If you can live with black and white, Megascan >(412 443-5820) is now marketing a 4096x3300 screen and controller which >work with ordinary ATs. Have you ever seen a Sun workstation? I have yet to see graphics on any PC match that. ANY PC. You might beat it in resolution, but you'll lose out on speed in generating the graphics. More on this later. >And, finally, for people who need to be able to >REALLY crunch numbers, do color rendering, ray-drawing etc., or simply >always secretely wished to have a personal computer with some reasonable >fraction of Cray power, there is the Fairchild Clipper set, 5-8 mips >sustained average, 30 mips peak, 2 megaflops etc., with at least one >organization selling a complete AT converted this way for less than >$8000 right now. OK. Now run something on your supersystem. You have described a $20,000 system (a couple of grand for the machine, a couple of grand for the graphics, 8 grand for the coprocessor, and of course you need a good hard disk). Now is it PC compatible? Bet not. Not 1/100th of the software available will run on it at anything better than normal performance. Your system is no longer in conformance with PC standards. Notice that an enhanced UNIX machine will generally take advantage of enhancements. Usually the compiler or the compiler's library have routines to test for these enhancements and take advantage of them, regardless of how nonstandard they are, and they are used when you compile the program. There is no demand for software that looks for your $8k coprocessor, so nobody makes it, so what good is the coprocessor. >Any way I look at it, it appears that the ultimate computer for most >people, the inexpensive AT clone, is already here, now, and I think most >users know this. True. They're certainly nice. But we really need a nicer operating system -- one which can make software ignore hardware differences and take full advantage of your system. Right now, UNIX is a good choice. I'm not saying it's the best choice, only that it's available. Personally, I would like a Mac-like interface, but with less mouse usage, with multitasking. Note that this, too, can be implemented under any UNIX-like operating system, in the same manner as software that would enable program suspension. >Ted Holden >HT Enterprises Sorry if I upset you with my flame, but it is my feeling that as machines become powerful enough to use better (yes, better!) operating systems, they should do so. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Jim Frost * The Madd Hacker | UUCP: ..!harvard!bu-cs!bucsb!madd H H | ARPA: madd@bucsb.bu.edu H-C-C-OH <- heehee +---------+---------------------------------- H H | "We are strangers in a world we never made"