Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!utgpu!water!watnot!watmath!clyde!rutgers!sri-spam!ames!amdahl!dlb!sun!cmcmanis From: cmcmanis@sun.UUCP Newsgroups: comp.sys.amiga Subject: Re: Some thoughts and questions about the Ranger Message-ID: <15020@sun.uucp> Date: Fri, 13-Mar-87 15:34:45 EST Article-I.D.: sun.15020 Posted: Fri Mar 13 15:34:45 1987 Date-Received: Sat, 14-Mar-87 11:41:18 EST References: <140@tahoma.UUCP> <543@dragon.tc.fluke.COM> <692@xanth.UUCP> Organization: Sun Microsystems, Inc. - Mtn View, CA Lines: 85 Keywords: clone antique mistake derivative hopeful "maybe next time" Summary: New Chips cost big bucks In article <692@xanth.UUCP>, Kent Paul Dolan writes: > In article <543@dragon.tc.fluke.COM> Kurt Guntheroth writes: > >>I would spend $1-2K for a computer for myself. I wouldn't spend $5K. > > I wouldn't either, but I would spend an extra $500 for an MMU, a 68020 or, > better, 68030 with a doubled clock speed, and a floating point accelerator, > with the software to make it work. Given the way Motorola costs things, I would say that samples of the 68030 will probably go for about $500 and be available to the general public in early '88. Boards that use them maybe in late '88 or early '89. Sheer speculation would put the cost of the '881 at about $100 at that time. So look to pay a lot more than $750 for a an add-on board with those two chips on it. (In 1989 of course) > ... The current OS make the fallacious > assumption that all "professional" software vendors are capable of writing > software that does its own memory management. This violates the standard > practice of doing a job that must be done everywhere, in one place instead, > and doing it right. Memory is just another "device", and there should be > one "driver" taking care of it, not one per program. The results of this > one execrable design decision has done more to hurt the Amiga's sales than > all other weak features put together. No business can live with a machine that > drops dead hourly. Worse yet, memory management is so simple that most CS > undergraduate programs include writing a memory management routine as a > class project in the data structures class. Why wasn't it done right in > AmigaDOS? With an MMU, perhaps the OS upgrade would be easy enough to tempt > Commmodore to do the job properly next time. Maybe I am not reading what Kent wrote, but from what I know of the Amiga it has a first rate memory management package. The routines AllocMem, FreeMem, AllocRemember, and FreeRemember provide excellent memory management support. I tell them how much memory I want, and what restrictions it has (must be CHIP, clear it first, must be PUBLIC etc) and they go out and get it for me, and tell me if it is unavailable. The only thing they don't do is prevent me from freeing someone elses memory or trying to free my own memory twice. If I don't mind the performance hit I can layer these routines through my own that make sure I haven't done one of these cardinal sins. >>An Amiga-as-a-UNIX-box will never sell. SUN already does that. Apple will >>be there soon. > > I disagree. A UNIX (tm) system at Amiga 1000 prices would sell like hotcakes > to universities, ... etc etc etc Remember when this was "If someone could sell a UNIX system for under $10K they would sell millions of them!" ? The point should be "Is Commodore the company to sell a cheap UNIX box?" I don't think so, not at this time of their life. > > However, AmigaDOS is the biggest sales weakness of an otherwise excellent > product... I would suggest that many people said "Multitasking O/S ? This is the machine for me!" I was one of them, there were others. AmigaDOS is not UNIX, it doesn't try to be UNIX, and in many cases is 'better' than UNIX for what it does. Have you ever tried to write fast real time response software under UNIX ? Not to start any flames here, it is *different* and does some things nicely, it also assumes your program won't beat up other programs. My favorite corallary is the 'C vs PASCAL' approach to O/Ses. Ever notice how the PASCAL advocates will point out that in C you can write all over the other variables in your program simply by exceeding the bounds of an array? That's because C expects you to declare an array of the size you need and not write out side of it. It is powerful and also dangerous. PASCAL will stop your program and tell you you tried to go out of bounds on an array. AmigaDOS and UNIX are like that, AmigaDOS assumes you know what you are doing and won't scribble over everyone elses memory, but if you want to you can. UNIX won't let you (unless you are the super user). Again, it is powerful and it is dangerous. > UUCP : kent@xanth.UUCP or ...seismo!decuac!edison!xanth!kent And interesting machine indeed. As the level of awareness as to just how flexible the Amiga OS is becomes prevalent, more and more of the 'problems' with the Amiga will be solved by innovative, compatible, and portable software. --Chuck McManis uucp: {anywhere}!sun!cmcmanis BIX: cmcmanis ARPAnet: cmcmanis@sun.com These opinions are my own and no one elses, but you knew that didn't you.