Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!unmvax!pprg.unm.edu!hc!lll-winken!uunet!tektronix!reed!mdr From: mdr@reed.UUCP (Mike Rutenberg) Newsgroups: comp.sys.next Subject: Re: NeXT alternatives Message-ID: <12236@reed.UUCP> Date: 30 Mar 89 07:44:05 GMT References: <12192@reed.UUCP> <700@adobe.UUCP> Reply-To: mdr@reed.UUCP (Mike Rutenberg) Organization: Reed College, Portland OR Lines: 113 What the heck - I feel like a little technical jousting (in good spirits because it was a wonderful day here today with sunshine and neat hail)... First I should indicate that I have the most respect for the what NeXT has managed to put together. They have taken a number of separate technologies, have blended them using their considerable talent and sweat into a neat little box. I want to write fun explorational educational physics simulations for the NeXt box (as well as a personal phone-mail system). But I don't have one because it is: (1) clearly not a machine for individuals (as NeXT will tell you) (2) not available, even to individuals who can afford it How I evaluate the machine is what interfaces or resources it provides me as a (1) programmer and (2) user. The former comes from NeXT, the later is what I can purchase from Egghead discount software in two or three years. In article <700@adobe.UUCP> greid@adobe.COM (Glenn Reid) writes: >Where is Display PostScript? Where is the "mainframe-on-a-chip" >architecture? Where is the Digital Librarian? Where is Mach? Display postscript is really quite nice, but as a programmer I'm not currently willing to pay $8000 for it. Most of the stuff I want to draw is easily expressed using a Macintosh's Quickdraw or in the Windows/PM API. Some of it would be a bit more uniform and fun in display Postscript, but then it *might* also be a little slower. The "mainframe-on-a-chip" architecture, as you so appropriately quote it, is apparently a big multiported dma chip (with some channel controller features?), and is a way that NeXT is choosing to optimize their IO. I won't see it as a programmer or as a user. As long as my machine has fast IO, be it an OS/2 box or whatever, I'm not too concerned that it has a "mainframe-on-a-chip" architecture. Some of the 386 boxes, suns, and other machines seem to be able to move things around pretty quickly. I would encourage somebody to benchmark a NeXT and a Sun/3 doing IO to some combination of optical, SCSI, and nfs "disks" so that it is will be easier to evaluate this feature. I suspect their implementation of the laser printer interface (requiring lots of bits from memory) is one of the things that makes this custom chip necessary. Mach is a cool kernel. OS/2 kernel looks *surprisingly* like it in some ways. I would be very surprised if you don't see similar features from Sun and Apple. Threads everyone?! Tell me more about the digital librarian. In what ways it it unique? How will it make me more happy? >The thing that impressed me the most about the NeXT machine was not the >surface-level features, but the complete system design from bottom to >top. Not only is there a DSP chip, but the system architecture is >strong enough to make it useful to you. It runs Mach (Unix), not >multi-finder or OS/2. It has Display PostScript built in on the ground >floor, not added as afterthought. Everything is carefully designed and >integrated. You can invoke Webster's dictionary from within a word >processor. You can play tunes while you edit. The "complete system design" of the NeXT box is very nice. And I agree that everything seems to be carefully designed and integrated. I disagree however that other systems are at any disadvantage because they were shipping while NeXT was maturing. Each design philosophy has different strengths. The DSP does not run Mach. Imagine an OS/2 based 386 box with an i860 coprocessor also running OS/2 (in a role similar to the DSP). That would be significantly faster (w/ floating point), more uniform to program, and could be added to the system by those people that really want it. The example is not an exact analog to the NeXT box but then the point is to do interesting things with computers rather than build NeXT clones. Also -- modular systems make it a lot easier to add something like display postscript and reduce the necessity of having had it on the CPU since day 1. It is not clear to me that the programmer's interface to NeXT display postscript need be any different than the interface to OS/2 display postscript. >Look at Microsoft Windows, which is a "feature imitation" of the >Macintosh interface. Is it as good? Is it integrated into the machine >itself? Do all programs run under it? I am not sure totally what to make of this. Not all "NeXT" programs run under NeXT step - especially sun binaries or anything which is pure unix. Similarly "Windows" has programs that run under it and programs that don't. >I think that you cannot simply copy what NeXT has done and run it on an >existing machine. Many of the good ideas visible in NeXT are not only ripe, but being picked by other companies. If I can get 90% of a NeXT box in a GOOD $600 software package, I will go for it. Assertion: NeXT and its "alternatives" are both systems of the future. Neither of them exist to any full degree, except in press releases, and both will mature in about two years. For 1989, most of us are stuck with DOS, MacOS, and type-at-it unix. I just wish my Mac Plus was getting quicker. Actually, one of the nice things about programming on the Mac Plus is you can be always reamazed at how much faster and more colorfully the program you are writing can run on a Mac II. Mike -- Mike Rutenberg Reed College, Portland Oregon (503)239-4434 (home) BITNET: mdr@reed.bitnet UUCP: uunet!tektronix!reed!mdr Note: My written word represents no known organization or person.