saustin@bbn.com (Steve Austin) (12/02/90)
I have been reading comp.os.minix for some time now and have been thinking of buying MINIX for my 386 IBM clone. I was just poised to oder it when I saw Glen Overby's "MINIX information sheet (1 Dec 1990)", which amongst other thing announced a price rise from $80 to $170. While I respect the aims of the MINIX users' group, I would like to ask some question (which I do not mean to be criticisms of Glen Overby, Michael Tannenbaum, the MINIX users' group or MINIX itself). 1) Why is Prentice-Hall jacking the price up and who is getting the money? While I realize that MINIX 1.5 is an improvement on earlier versions, it seems to me that it is partially improved by the users group. So who is getting the extra money? I would hope that it is the programmers who contribute to MINIX, but if (as I suspect) Prenitce-Hall is trying to milk what it thinks is a cash cow, they can take a walk. There are other, cheaper, versions of UNIX for the PC. 2) What is going to happen to the price when MINIX gets virtual memory? It seems to me that various dedicated souls are working on implementing VM for free. Can I expect to see another price hike when this happens. If so, who gets the money? 3) On a less emotional note, what is the max size of program you can run on MINIX for a 386. If it runs in the same mode as an 8086, I suppose the answer is 640k. Since Glen's information sheet mentionned that people are working on 32bit MINIX for the 386, I suppose that means that the 386 operates in brain-dead mode. 4) I notice that a ANSI C compiler is being developed. However, somebody has recently announced that GNU's gcc works for 386 MINIX. Also, gcc has been producing 68k code for a long time. Wouldn't it be easier to work with GNU to get gcc, gdb, g++ etc integrated into the MINIX environment. Again, no offence to MINIX intended, I am basically wondering whether history is repeating itself with MINIX. Steve Austin
paradis@acestes.UUCP (Jim Paradis) (12/02/90)
In article <61262@bbn.BBN.COM> saustin@bbn.com (Steve Austin) writes: >1) Why is Prentice-Hall jacking the price up and who is getting the money? > While I realize that MINIX 1.5 is an improvement on earlier > versions, it seems to me that it is partially improved by the users > group. So who is getting the extra money? I would hope that it is the > programmers who contribute to MINIX, but if (as I suspect) > Prenitce-Hall is trying to milk what it thinks is a cash cow, they > can take a walk. There are other, cheaper, versions of UNIX for the > PC. The reason for the price increase is that the package got bigger and better. The $80 price when MINIX first came out got you a box of eight floppies. Period. No documentation, nothing. If you wanted documentation, you bought "The Book" (another $40). The new $170 price reflects the additional costs of producing a package of seventeen floppies plus documentation. P-H isn't just "milking this like a cash cow"; they're offering significantly more value, and adjusting the price accordingly. BTW, what "other, cheaper versions of UNIX for the PC" did you have in mind? Certainly none of the SysV derivatives; they're all mucho $$$. PCNX? A total joke, even though it does come with source. Coherent? Nice as far as it goes, but no source. Face it, MINIX is the only OS out there which offers full source AND a vibrant user community which is continually making new additions. In fact, thanks to that user community, I was able to keep my investment in MINIX down to my original $80: Starting with the 1.1 disks which I had on my shelf from a long time ago, I upgraded to 1.2, then 1.3, then 1.5, then to Bruce's 386 version. With the help of numerous other packages (freely offered by members of this user community), I am able to post this article to you from my home MINIX machine. >2) What is going to happen to the price when MINIX gets virtual memory? > It seems to me that various dedicated souls are working on > implementing VM for free. Can I expect to see another price hike > when this happens. If so, who gets the money? Depends. If VM and 386 support are included in 1.6 and the additional support requires significantly more disks or documentation, then yes you may see a price hike... but (IMHO) ONLY if the additional media cost warrants it. From what I've seen of MINIX pricing trends, it has always been priced fairly to reflect the amount of material supplied. I expect this trend to continue. >3) On a less emotional note, what is the max size of program you can > run on MINIX for a 386. If it runs in the same mode as an 8086, > I suppose the answer is 640k. Since Glen's information sheet > mentionned that people are working on 32bit MINIX for the 386, > I suppose that means that the 386 operates in brain-dead mode. If you use 1.5 out of the box, then the biggest program you can run on a 386 (or a 286 or an 8088) is 64K code/64K data. If you run in 32-bit mode with Bruce Evans' patches, then you're limited only by the physical memory size of your machine and the other processes running. If I can get my VM implementation working, then you'll only be constrained by the size of your hard drive 8-). >4) I notice that a ANSI C compiler is being developed. However, > somebody has recently announced that GNU's gcc works for > 386 MINIX. Also, gcc has been producing 68k code for a long time. > Wouldn't it be easier to work with GNU to get gcc, gdb, g++ etc > integrated into the MINIX environment. The trouble with GNU-anything is that it takes up a LOT of memory. I'm currently running "bash" (GNU's "bourne-again shell") as my default shell. If I use gcc to compile anything bigger than hello.c, then I'd better do it under /bin/sh instead because I'll run out of memory otherwise. And I'm running on a 4Mb machine. The ANSI C compiler is likely to be significantly smaller. >Again, no offence to MINIX intended, I am basically wondering whether >history is repeating itself with MINIX. Yes, history is repeating itself, but that's not a bad thing. At least as far as pricing is concerned, P-H has been eminently fair. There HAVE been complaints about their efficiency in SHIPPING, but that's another story...
ast@cs.vu.nl (Andy Tanenbaum) (12/02/90)
In article <61262@bbn.BBN.COM> saustin@bbn.com (Steve Austin) writes: >I would like to ask some question (which I do >not mean to be criticisms of Glen Overby, Michael Tannenbaum, the MINIX users' >group or MINIX itself). My brother's name is Mitchell, not Michael. >1) Why is Prentice-Hall jacking the price up and who is getting the money? The 1.5 distribution has 17 disks and an 688 page manual. It costs $169. The original 1.1 version had 8 disks and no manual and cost $79. There is a certain linearity there. A substantial amount of the price is related to the manufacturing and distribution costs (not to mention advertising and general corporate overhead). While it is true that a blank disk costs under a dollar, the cost of producing a color box with 17 disks full of information, a manual, a binder, etc. plus the cost of people who take orders etc is nonzero. One of the reasons Coherent was cheaper was their special introductory price (now raised, I believe), plus the fact that they use 1.2M disks and distribute binary only, so they can get away with only 4 disks. This excludes PCs and XTs. I have made a deliberate decision to stick with 360K disks to keep supporting XTs for the moment. Having two versions (XT and AT) in the past caused endless problems. How much does full XENIX, with C compiler, cost these days? > So who is getting the extra money? I would hope that it is the > programmers who contribute to MINIX, but if (as I suspect) > Prenitce-Hall is trying to milk what it thinks is a cash cow, they > can take a walk. There are other, cheaper, versions of UNIX for the > PC. P-H is making a smaller profit on MINIX than they make on books. They tend to regard MINIX sort of the way they regard Solutions Manuals that go with textbooks--as an extra added attraction to sell books. As to other, cheaper versions of UNIX, the only one I know of is Coherent, which I think is now $159 (but check this). For this $10 difference you get all the source code. Furthermore, I have posted all the upgrades to the net, and will continue to do so in the future, so everyone can upgrade for free indefinitely. >2) What is going to happen to the price when MINIX gets virtual memory? > It seems to me that various dedicated souls are working on > implementing VM for free. Can I expect to see another price hike > when this happens. If so, who gets the money? If VM goes in and more disks are needed, the price will invariably go up. However, perhaps I should state once more that which I have stated about n factorial times in the past. It isn't like I am sitting back on my rear end waiting for all these free software to appear and then greedily grabbing it. Rather the opposite. I don't want 32-bit mode. I don't want virtual memory. I don't want swapping. I don't want multiple consoles. I don't want X windows. I don't want uucp. I don't want mail. I don't want news. I don't want Berkeley anything. etc. etc. etc. etc. etc. etc. etc. etc. etc. I have been actively resisting all this stuff, as everyone who has been reading this newsgroup well knows. I even once categorically stated that I wouldn't accept protected mode, but the hue and cry from everybody made me eat crow and take it. All I really want is a simple 16-bit MINIX that conforms to P1003.1 and P1003.2 which students can fully understand. If history is any guide, I will probably end up taking some of the above stuff if put under enough pressure from the user community, but grudgingly at best. I find it slightly unfair to be accused of profiteering by taking all this stuff when actually I am being force fed it rather against my will. >3) On a less emotional note, what is the max size of program you can > run on MINIX for a 386. If it runs in the same mode as an 8086, > I suppose the answer is 640k. Nope. The maximum size for programs is 64K + 64K. This keeps X windows at bay. >4) I notice that a ANSI C compiler is being developed. However, > somebody has recently announced that GNU's gcc works for > 386 MINIX. Also, gcc has been producing 68k code for a long time. > Wouldn't it be easier to work with GNU to get gcc, gdb, g++ etc > integrated into the MINIX environment. GNU is huge and won't run in 64K. Furthermore, it has legal restrictions that certainly add complications at the very least. Besides, MINIX was never intended as a GNU clone. The ACK ANSI C compiler is now in beta test. >Again, no offence to MINIX intended, I am basically wondering whether >history is repeating itself with MINIX. It probably is. I regard MINIX as the competitor and possibly successor to XINU. It is aimed at the same audience as XINU. To me, success or failure is measured in some sense in how well it does compared to XINU. Andy Tanenbaum (ast@cs.vu.nl)
paradis@acestes.UUCP (Jim Paradis) (12/03/90)
You took the words right outta my keyboard, Andy 8-) Actually, as for 32-bit and VM and all that... I for one agree that the official release of MINIX should be "lean and mean", and that you should resist creeping featureism as much as possible. I have no problem with adding additional features (like the VM implementation I'm working on) as long as they remain unofficial additions to the base release and not a part of the base system itself. Bruce Evans, Earl Chew, et al have shown us that "unofficial" need not mean "unprofessional". Perhaps one way to "resist" the pressure to add features would be to do what a lot of other OS vendors do: have an "official" distribution, and also have an "unspported features" tape/disk set. The latter would contain all the goodies people are likely to ask for but that you don't want to make part of the "official" distribution. It comes on a strictly as-is, no-support, buyer-beware product (sort of like a lot of anonymous-ftp software is today). It's a thought, and it's a way to keep the wolves at bay 8-) -- Jim Paradis UUCP: harvard!m2c!jjmhome!acestes!paradis 9 Carlstad St. AT&T: (508) 792-3810 Worcester, MA 01607-1569 ICBM: 42deg 13' 52", 71deg 47' 51"
saustin@bbn.com (Steve Austin) (12/03/90)
ast@cs.vu.nl (Andy Tanenbaum) writes: >I find it slightly unfair to be accused of profiteering by taking all this >stuff when actually I am being force fed it rather against my will. I replied to Andy via e-mail, but I think it is also appropriate to post a public apology, if I gave the impression that I was accusing Andy of profiteering from MINIX. The change in price of the distribution discs ($80->$170) caused me to worry that PH was trying to increase its profit from MINIX, but Andy's reply has proved these fears ill founded. I did not intend to criticise Andy's motives (or those of any of his family). I'm sure that anybody who has read his OS book or who reads this newsgroup has nothing but the highest regard for Andy's ability or motives. Steve Austin
HBO043%DJUKFA11.BITNET@cunyvm.cuny.edu (Christoph van Wuellen) (12/03/90)
It it prefectly legal to upgrade from an old version using postings from the net. They say the price increase correlates well with the number of floppy disks that are in the distribution.
nall@sun8.scri.fsu.edu (John Nall) (12/03/90)
In article <61275@bbn.BBN.COM> saustin@bbn.com (Steve Austin) writes: >ast@cs.vu.nl (Andy Tanenbaum) writes: > >>I find it slightly unfair to be accused of profiteering by taking all this >>stuff when actually I am being force fed it rather against my will. > >I replied to Andy via e-mail, but I think it is also appropriate to >post a public apology, if I gave the impression that I was accusing >Andy of profiteering from MINIX. The change in price of the >distribution discs ($80->$170) caused me to worry that PH was trying to >increase its profit from MINIX, but Andy's reply has proved these fears >ill founded. I did not intend to criticise Andy's motives (or those of >any of his family). I'm sure that anybody who has read his OS book or >who reads this newsgroup has nothing but the highest regard for Andy's >ability or motives. > > Steve Austin Not only do we have the highest regard for his ability *and* motives, but we also are completely aware as to the reason for the P-H price hike. The reason we are completely aware is because Andy posted it to the net a long time ago. We are also aware that since one has to buy Minix only once (I bought 1.1 a long, long time ago) and after that it can be kept upgraded at little or no cost (depending on individual network costs) it is pretty damn cheap. We are also used to people making snide comments which indicates they do not know what they are talking about. Some of them later prove to be pretty good people, so the apology is accepted in the spirit in which it was offered (yes, some of us found it extremely offensive). Welcome to the net. Cheers, -- John W. Nall | Supercomputer Computations Research Institute nall@sun8.scri.fsu.edu | Florida State University, Tallahassee, FL 32306 WB4LOQ (why? I dunno....everyone else seems to be doing it. _._)
mat@emcard.UUCP (W Mat Waites) (12/03/90)
In article <8395@star.cs.vu.nl> ast@cs.vu.nl (Andy Tanenbaum) writes: >How much >does full XENIX, with C compiler, cost these days? It would be more fair (from a quality, support and program size point of view) to compare to Coherent, don't you think???? ($99.95) Mat -- W Mat Waites | Walking the wire is living, {gatech,emory}!emcard!mat | the rest is just waiting. - Wallenda
owens@acsu.buffalo.edu (bill owens) (12/03/90)
ast@cs.vu.nl (Andy Tanenbaum) writes: >I regard MINIX as the competitor and possibly successor to >XINU. It is aimed at the same audience as XINU. To me, success or failure is >measured in some sense in how well it does compared to XINU. If we let history be the judge, the 266 articles which have arrived here from comp.os.xinu compare poorly against the 13532 articles for comp.os.minix. I've yet to hear of any use of xinu beyond the purely academic; my inquiries about it were met with immediate, gracious responses from Dr. Comer, but his information only showed it to be a non-contender. And to make the point that Minix is still a viable teaching system, my instructor for OS class next semester was only willing to consider it after he found out what it actually contained; he was afraid it wouldn't have enough of the features of a 'real-world' OS. His viewpoint was that a 'toy' system would make it harder to teach some of the most important parts of practical OS work, since it wouldn't include them. We'll still have to study VM and some other esoteric things without real examples in code (except for BSD :), but the class as a whole will be taught using Minix 1.5.10. Bill.
rterry@hpcuhc.cup.hp.com (Ray Terry) (12/04/90)
> Not only do we have the highest regard for his ability *and* motives, but > we also are completely aware as to the reason for the P-H price hike. The > reason we are completely aware is because Andy posted it to the net a > long time ago. We are also aware that since one has to buy Minix only > once (I bought 1.1 a long, long time ago) and after that it can be kept > upgraded at little or no cost (depending on individual network costs) it > is pretty damn cheap. > > We are also used to people making snide comments which indicates they do > not know what they are talking about. Some of them later prove to be pretty > good people, so the apology is accepted in the spirit in which it was > offered (yes, some of us found it extremely offensive). Welcome to the net. Lots of we's and us's here. Got a mouse in your pocket? [grin] I, for one, read the original basenote and thought it not offensive. Just seemed like reasonable questions. Nothing wrong with asking them... Ray
a577@mindlink.UUCP (Curt Sampson) (12/04/90)
> mat@emcard.UUCP writes: > > In article <8395@star.cs.vu.nl> ast@cs.vu.nl (Andy Tanenbaum) writes: > >How much > >does full XENIX, with C compiler, cost these days? > > It would be more fair (from a quality, support and program size point of > view) to compare to Coherent, don't you think???? ($99.95) I think so. I've been using Coherent for a week or so now. A friend is loaning me his copy of Minix while he goes away for Christmas, so I'll drop that into another partition and do a head-to-head comparison. If anybody cares, I'll mail or post my opinions. From what I've read here and my playing about with Coherent I'd say that they're rather differently oriented. I've not yet been able to crash Coherent (even by doing nasty things like filling up the process table). I've found a few things that are definite problems, though they can be worked around. Coherent has some nifty features, such as the /dev/ram driver and a sixteen bit compress. On the other hand, I really wish I had the source to ls so that I could fix it to interpret the second letter of the command name as an option. I also wish that their manual pages were in a normal man format rather than the Coherent Lexicon format. I wish I had a gotten a lot more technical documentation with the thing (though I gather I can write away to get the goods on writing device drivers and the like). In short, it's definitely got some nifty features, it's well supported, has few bugs, and is cheap. There's supposed to be a large model compiler coming out soon, too. But it's not very BSDish, and still no source code... cjs -- Curt_Sampson@mindlink.UUCP {uunet|ubc-cs}!van-bc!cynic!curt curt@cynic.wimsey.bc.ca
tim%maths.tcd.ie@pucc.princeton.edu (Timothy Murphy) (12/06/90)
> In article <61275@bbn.BBN.COM> saustin@bbn.com (Steve Austin) writes: > We are also aware that since one has to buy Minix only > once (I bought 1.1 a long, long time ago) and after that it can be kept > upgraded at little or no cost (depending on individual network costs) it > is pretty damn cheap. I'd be really interested to know the answer to the following query: if one gets *one* version of Minix (say for the PC), can one modify that to run on the Mac, say. Is it possible? Do there exist diffs for 80x6 -> 680y0 Minix? I asked on the net some days ago, but didn't get any response. I'd be really grateful if someone could let me know -- I'm hoping to start an Irish Minix Users Group, and would like to demonstrate Mac and ST versions, as well as PC. But forking out $169 x 3 is a bit steep. Timothy Murphy e-mail: tim@maths.tcd.ie tel: +353-1-772941 x 1507 s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland
awb@aipna.ed.ac.uk (Alan W Black) (12/07/90)
In article <61262@bbn.BBN.COM> saustin@bbn.com (Steve Austin) writes: >1) Why is Prentice-Hall jacking the price up and who is getting the money? > While I realize that MINIX 1.5 is an improvement on earlier > versions, it seems to me that it is partially improved by the users > group. So who is getting the extra money? I would hope that it is the > programmers who contribute to MINIX, but if (as I suspect) > Prenitce-Hall is trying to milk what it thinks is a cash cow, they > can take a walk. There are other, cheaper, versions of UNIX for the > PC. > The way I understand it is that the 80$ was for the base 1.3 system on 11 floppies while Minix 1.5 is on 20 floppies and includes a full manual (~700 pages). Thus as far as I see it the extra money goes on actual physical objects rather than as profit. I don't suggest that Prentice hall does not make a profit from this but the increase in price is due mostly to it consisting of more disks and a manual. >2) What is going to happen to the price when MINIX gets virtual memory? > It seems to me that various dedicated souls are working on > implementing VM for free. Can I expect to see another price hike > when this happens. If so, who gets the money? > If Minix grows again, say at version 2.0 I suspect that it will need more disks. I would personally suggest it should be distributed on greater density disks to save some money. However the important thing to me about Minix is that it comes with source. As yet there is *no* alternative anywhere near the price range of Minix that also offers source (without a AT&T source lincece). Also I would like to add that Andy Tanenbaum has indicated that he is not too happy about "big" extensions being added to Minix as he sees the system as a Teaching system for OS courses. It is possible that things like VM will never be included in the standard distribution and only be available as diffs which interested people can add. I also think that some of the pressure among Minix hackers to make Minix into a bigger unix (** I think it is already a *real* unix **) by adding VM, streams, graphics will die down when other "free" unix systems come out. GNU (the FSF) intend to bring out a free unix based on CMU's Mach, and Berkeley are working on a 4.4 "Detox" which will contain no AT&T code and hence be "free" for anyone. This I think wont happen for at least a year but should happen in under two. Also note that I say "free" in that the software will be free but it will still cost money for media and manuals. I doubt these will cost less than $200. (Though they should be ftp'able). >3) On a less emotional note, what is the max size of program you can > run on MINIX for a 386. If it runs in the same mode as an 8086, > I suppose the answer is 640k. Since Glen's information sheet > mentionned that people are working on 32bit MINIX for the 386, > I suppose that means that the 386 operates in brain-dead mode. > The major 386 patch that is used is Bruce Evans port (There is another recently announced too.) Both these do not restrict the size of processes (well they do actually but to somewhere around 4Gigabytes). I run processes of over 2Megabyte on my machine with no problems (except I need more physical memory :-) >4) I notice that a ANSI C compiler is being developed. However, > somebody has recently announced that GNU's gcc works for > 386 MINIX. Also, gcc has been producing 68k code for a long time. > Wouldn't it be easier to work with GNU to get gcc, gdb, g++ etc > integrated into the MINIX environment. > GNU as a project has very little money but depends largely on the support of people on the net. When ports are done for Minix of GNU software, information does go back to the FSF. The GNU sofwtare has not yet been POSIX'ified but probably will be with help from Minix people (Bruce Evans has already done this). As for other ports like my own gcc and gas, we did find bugs in gas that have since been corrected and communicated with Richard Stallman about how best to do floating point in gcc. Most of our port was in the problems of boot strapping a gcc. The diffs included are pretty minimal. I have had to change more in gnu software to get it run on some so-called supported systems (or perhaps I should say so-called unix systems :-) than is required for some things under Minix. I see the porting of GNU stuff to Minix as directly helping GNU. As for ANSI C compilers, I think Minix can have its own and need not require gcc. Gcc is very big, and needs lots of memory to run and if we are already worried about the cost of distribution it is cheaper to have a smaller ANSI C compiler. >Again, no offence to MINIX intended, I am basically wondering whether >history is repeating itself with MINIX. > > Steve Austin No offense taken, as for history repeating itself, I thought that was the whole point :-) hope this helps to explain Alan Alan W Black 80 South Bridge, Edinburgh, UK Dept of Artificial Intelligence tel: (+44) -31 225 7774 x228 or x223 University of Edinburgh email: awb@ed.ac.uk
feustel@netcom.UUCP (David Feustel) (12/07/90)
I'm glad that netpeople continue to upgrade minix and make their upgrades available for free on the net. I'm also happy to pay PH for making a convenient package out of all the upgrades so that I don't have to do as much work to get the upgraded system working on my machine. -- David Feustel, 1930 Curdes Ave, Fort Wayne, IN 46805, (219) 482-9631 EMAIL: netcom.uucp